Release Notes

Spicerack Changelog

v8.16.2 (2024-11-18)

Minor improvements

• redfish: add response logging for request() to better capture errors that are hard to reproduce.

Bug fixes

• mysql_legacy: improve DRY-RUN support in execute() and documentation for it on the other methods.

v8.16.1 (2024-11-14)

Bug fixes

• mysql_legacy: fix set_master_use_gtid() query, its value it's part of the syntax, avoid pymysql quoting it.

• mysql_legacy: fix query formatting in set_replication_parameters().

• mysql_legacy: fix check in replication_lag() that would raise if the lag is 0.0s.

• doc: fix example code bug missing a reference to self.

v8.16.0 (2024-11-13)

New features

• mysql_legacy: add MysqlClient class as a copy of the mysql.Mysql class to later merge those two modules together.

• mysql_legacy: improve pymysql usability adding some new helper methods:

  • execute(): to execute a query that doesn't return anything via pymysql.

  • fetch_one_row(): to execute a query with pymysql that should return one row and return it.

  • check_warnings(): to check if in the last statement there was any warning raised and ask the user what to do.

• mysql_legacy: in the Instance class convert all internal queries to use the new methods to use pymysql instead of executing queries via ssh.

Bug fixes

• mysql: remove deprecated call to query() method of pymysql that is for internal use only. Convert it to a cursor().execute() call that is the part of the public facing API.

v8.15.2 (2024-10-31)

API breaking changes

• elasticsearch: removed ElasticsearchHosts.get_remote_hosts() getter, superseded by the new RemoteHostsAdapter.remote_hosts().

• puppet: removed PuppetServer.server_host() and PuppetMaster.master_host() getters, superseeded by the new RemoteHostsAdapter.remote_hosts().

• Because of the very low usage of the above methods this didn't warrant a major release. Reporting it as breaking here for completeness, their usage will be fixed right after releasing this version.

Minor improvements

• remote: add remote_hosts getter to the RemoteHostsAdapter to ease the use from clients. This also removes one-off getter from other classes in the puppet and elasticsearch_cluster modules.

Bug fixes

• orchestrator: do not retry on 500s as orchestrator tends to reply to non-existing objects with a 500 with a JSON response, do not retry the request.

• mysql_legacy: accept any exit code for systemctl status to prevent having RemoteExecutionError exceptions.

• mysql_legacy: add getter for the Instance's socket property.

• mysql_legacy: fix list_host_instances() detection of single and multi-instances independently of the status of the systemd unit.

v8.15.1 (2024-10-23)

Bug fixes

• orchestrator: fix bug with older requests that doesn't have the JSONDecodeError exception.

• service: change depool_threshold field to float following Puppet related change.

v8.15.0 (2024-10-23)

API breaking changes

• mysql: refactor this currently unused module to be up to date with the current infrastructure while simplifying it. Because of the unused nature of the module this didn't warrant a major release. Reporting it as breaking here for completeness.

New features

• orchestrator: add a new module to interact with Orchestrator's APIs.

• apiclient: add a generic API client module and related Spicerack accessor.

Minor improvements

• redfish: use the new apiclient module.

• redfish: add UEFI functions to check if a host is setup with UEFI and to boot into UEFI HTTP.

• puppet: add format option to hiera_lookup.

• mysql_legacy: add data directory accessor.

• mysql_legacy: re-order the CORE_SECTIONS constant from the less impactful to most impactful.

• mysql_legacy: get systemd status for instance to easily check if the instance is running or not.

• mysql_legacy: add cursor method to the Instance class to get a mysql client connection to the instance.

• remote: add dry_run getter for RemoteHosts, useful for RemoteHostsAdapter implementations.

Bug fixes

• dhcp: Add option to omit sending filename to a vendor, used for the Debian Installer.

Miscellanea

• doc: removed deprecated call to sphinx_rtd_theme.

• tox: only install flake8 when running flake8.

• tests: fix issues reported by pylint >3 and pin Prospector.

v8.14.0 (2024-09-30)

New features

• dbctl: add new module to interact with dbctl (T362893).

  • Add a new spicerack accessor to get a Dbctl instance.

  • From the Dbctl instance allow to access the dbctl libraries for Instance, Section and DbConfig (mediawiki config).

  • Dry-run support is ensured via the parent Confctl class that sets the read_only argument to the ConftoolClient instance accordingly.

Minor improvements

• confctl: add native support for RO in conftool

  • The spicerack interface to Conftool via the ConftoolEntity class does honor dry-run itself, although conftool was not having a dry-run support.

  • With recent contool development we can now use ConftoolClient to initialize it and this interface allows to set a read_only parameter.

  • The ConftoolClient interface abstracts the setup of the conftool client from the caller, in place of the to-be deprecated kvobject.KVObject.setup method.

  • Use the read_only parameter when in dry-run mode, both for safety reasons and also to enable using more complex conftool operations, such as the ones offered by the dbconfig extension.

Miscellanea

• netbox: removed Netbox 3 backward compatibility, all existing Netbox instances are 4+.

v8.13.1 (2024-09-17)

Bug fixes

• mysql_legacy: Add a 1 second sleep after start_slave() to ensure that a subsequent call to show_slave_status() would be reliable. Rename master_use_gtid() to set_master_use_gtid() for better clarity of the RW nature of it.

v8.13.0 (2024-09-06)

Minor improvements

• doc: add intersphinx_timeout (T367410).

  • The config should allow to have quicker Debian builds when the network is not available.

• redfish: allow 200 responses in chassis_reset (T365372).

  • On Supermicro nodes, chassis_reset's HTTP call gets a HTTP 200 from the BMC, not 204. It seems ok to relax the condition and allow both 204 and 200, without extra logging since the Supermicro's BMC response is not useful.

• redfish: catch no-json-responses in change_user_password (T365372).

  • The Supermicro's Redfish implementation works the same as Dell's in change_user_password, except for the fact that no JSON response is returned.

• redfish: introduce the AccountManager URI for DELL (T365372).

  • From various tests it seems that the /redfish/v1/AccountService URI works on DELL too, but only for "read-only", namely getting accounts' info. Refactor a bit the redfish class and the find_account() method to take this into account.

v8.12.0 (2024-09-02)

Dependencies breaking changes

• setup.py: update pynetbox to 7.4 (T373794).

  ** After T371890#10081172 Spicerack fails to build due to pynetbox,

  since it was upgraded to 7.4

v8.11.0 (2024-09-02)

New features

• dhcp: allow empty distro for DHCPConfMac and DHCPConfOpt82 (T365372).

  • Allow "distro" to be empty, so that the correspondent pathprefix config is not rendered. This is useful when we want to add DHCP configs for IP configuration only, like the Supermicro BMC/mgmt interface.

Minor improvements

• tox: run less environments on CI (T372485).

v8.10.0 (2024-08-01)

New features

• mysql_legacy: Instance class improvements (T371351).

  • Rename use_gtid() to master_use_gtid() to follow MySQL naming convention. Change its signature to accept a setting parameter to pick which valid value to use.

  • Introduce a MasterUseGTID enum class to represent the valid values that can be used for the MASTER_USE_GTID parameter.

  • Add a run_vertical_query() method to run a query with the vertical output format (G) and parse its result to a list of dictionaries.

  • Adapt the other methods that would benefit of the above method to use it.

• redfish: add the add_account function (T365372).

  • Supermicro ships their servers with the BMC admin account set to ADMIN, meanwhile we standardized the usage of root inside Wikimedia (basically what Dell does by default). Added a new add_account function that uses Redfish to create a new account.

v8.9.0 (2024-07-25)

New features

• dhcp: add dhcp_filename and dhcp_options for DHCPConfMac and DHCPConfOpt82 (T363576).

  • The DHCP configuration can now be customized with ad-hoc filename and DHCP option settings.

Bug fixes

• mysql_legacy: fix Instance's upgrade path (T367496)

  • The binary that runs the mysql upgrade needs to run other tools within the same directory and when called with a full path it will try to run them from the same path. But because the mysql_upgrade binary has a chain of symlink, we need to resolve them first before being able to run it with the full path.

v8.8.0 (2024-07-18)

New features

• netbox: add support for Netbox 4 (T336275).

  • Limited support for cables with multiple terminations per sides: the first termination is the only one considered.

Minor improvements

• netbox: refactor tests to be more flexible, and adapt them for Netbox 4.

v8.7.0 (2024-07-16)

New features

• redfish: add property for storage manager URI (T365372):

  • Add a new property for RedfishDell and RedfishSupermicro to be used as helper in various cookbook that require the URI path to get Storage Members info.

Minor improvements

• redfish: simplify interface of Redfish classes (T365372):

  • Now that we have two implementation we can see the common parts and simplify a bit the hardcoded bits in both derived classes of the Redfish class.

  • Define only the specific service name, not the whole path in the concrete classes and define the path in the parent class.

  • Define the service names as class properties instead of instance properties to reduce the number of lines and make it more readable, we don't really need the strictness of inheritance to ensure we add all of them when implementing a new vendor, it's fairly rare.

• mediawiki: update siteinfo URL to use mw-api-int (T367949)

• mysql_legacy: update core sections (T367496):

  • The external storage sections were recently rotated to new ones.

Bug fixes

• mariadb: bugfixes mysql_legacy (T367496):

  • We introduced a number of bugs in spicerack 8.6.0 that needs to be handled for automation implementations to begin.

  • Refactored and simplified a bit the new APIs.

  • Added full test coverage.

v8.6.0 (2024-06-12)

New features

• redfish: expand support for Supermicro hosts (T365372):

  • Allow RedfishSupermicro to be picked up in __init__.py based on what Netbox returns as manifacturer (and not just default to RedfishDell). Update tests to reflect this new behavior.

  • Move get_power_state() to an abstract method, to be implemented in vendor-specific classes. Update also tests to reflect this.

• mysql_legacy: improve support for MariaDB instances on each host (T343674).

Miscellanea

• redfish: fix typo in DellSCP's class description.

v8.5.0 (2024-04-15)

Minor improvements

• netbox: add functions to get and set the device name.

• elasticsearch: remove the dependency from elasticsearch-curator making the calls directly via the elasticsearch library (T345337 and T361647).

• alertmanager: add multi-instance and authentication support (T360932):

  • Add support for multiple alertmanager instances based on a configuration file. One of those instances can be marked as default which is used when the call to the Spicerack.alertmanager() or Spicerack.alertmanager_hosts() API is used without specifying a specific instance or some other API (like Service.downtime()) that does not support multiple instances is used.

  • Add support for per-instance HTTP basic authentication. The metricsinfra Alertmanager instance will be behind HTTP basic authentication to avoid exposing the read-write API to the entire wikiprod network (via the HTTP proxies). This patch adds support for configuring a username and a password to use on a specific Alertmanager instance.

Bug fixes

• puppet: make PuppetServer.destroy() have the same behaviour of PuppetMaster.destroy() and do not raise an exception if the host certificate is already missing (T360293).

Miscellanea

• setup.py: remove dependency elasticsearch-curator not needed anymore and remove upper bound for black linter that was there for incompatibilities with elasticsearch-curator.

• k8s: Remove use of @staticmethod in tests.

• tests: fix typos in tests that were erroneously calling mock methods with the wrong names.

• utils: remove --apply from isort's call in format-code, now the default in v5.

v8.4.1 (2024-03-06)

Minor improvements

• k8s: add getter for the Batch API.

v8.4.0 (2024-02-27)

New features

• netbox: allow to execute a Netbox script and retrieve the results.

• netbox: add getter/setter for primary IPs and access vlan.

Minor improvements

• ganeti: pass the v4 and v6 IPs to the VM as fw_cfg in the create command.

v8.3.0 (2024-01-29)

Minor improvements

• ganeti: add support for routed Ganeti (T300152).

Bug fixes

• alertmanager: fix timezone bug when run from a non-UTC computer (T347490).

Miscellanea

• setup.py: add missing classifier for Python 3.11.

v8.2.0 (2023-11-22)

Minor improvements

• puppet: add a hiera_lookup() method to the PuppetServer and PuppetMaster classes to perform a hiera lookup of a specific key from the perspective of a specific host.

v8.1.0 (2023-11-20)

Minor improvements

• remote: add a new RemoteHost.get_subset() method return a new RemoteHosts instance with a subset of the hosts. Useful when working with instances that inherit from RemoteHostsAdapter to be able to work on a subset of the hosts.

• service: Add ipip_encapsulation field to ServiceLVS to follow what's in Puppet.

• puppet: Update get_ca_server to also support SRV discovery records.

v8.0.3 (2023-11-16)

Bug fixes

• puppet: for the Puppet 7 migration set temporarily the return value of get_puppet_ca_hostname() hardcoded to puppetmaster1001 to allow to migrate the cumin hosts to Puppet 7.

Miscellanea

• doc: expand distributed locking docs, add an example of logging when unable to acquire a lock.

• spicerack: log at debug level some stats of each cookbook execution in a machine-readable format. This can be useful to generate some stats of the cookbook executions allowing to split them by exit code too.

v8.0.2 (2023-10-18)

Bug fixes

• locking: delete the key on etcd if no locks remain to keep etcd clean and avoid to left a lot of keys with emty dictionaries as values (T341973).

v8.0.1 (2023-10-18)

Bug fixes

• locking: fix path for Spicerack modules locks that was not correctly calculated.

v8.0.0 (2023-10-17)

API breaking changes

• dhcp: the spicerack.Spicerack.dhcp() accessor has changed signature and now accepts just a datacenter name instead of RemoteHosts instance. All cookbooks using this accessor had the same logic implemented to find the specific dhcp hosts in a given datacenter and this logic has been moved inside the accessor. All existing usage will be migrated at deploy time.

• netbox: remove methods fetch_host_status, fetch_host_detail and put_host_status that were deprecated since v0.0.50 and replaced by the spicerack.netbox.NetboxServer class. Some private methods have also been renamed to follow more closely Netbox namings.

New features

• Distributed locking support (T341973):

  • See the dedicated Distributed locking section of the documentation for a general overview.

  • Cookbooks class API additions to the spicerack.cookbook.CookbookRunnerBase base class:

    • max_concurrency class property to statically set the maximum number of concurrent runs of a given cookbook, enforced by the distributed lock.

    • lock_ttl class property to statically set the TTL of the distributed lock acquired for each cookbook run.

    • lock_args instance property to dynamically modify the locking arguments, for example based on the CLI arguments (RO vs RW mode of operations).

  • Cookbooks module API additions:

    • MAX_CONCURRENCY module constant to statically set the maximum number of concurrent runs of a given cookbook, enforced by the distributed lock.

    • LOCK_TTL module constant to statically set the TTL of the distributed lock acquired for each cookbook run.

  • Automatically acquire a lock for each cookbook run according to the values defined above.

  • spicerack: add a _spicerack_lock private accessor to get a lock instance to be passed to the Spicerack modules that would need to acquire a distributed lock with concurrency and TTL. It is different from the public accessor for the cookbooks because the key prefix is different to keep cookbooks custom locks separate from the spicerack modules ones. It's mentioned here as information for Spicerack developers.

Minor improvements

• dhcp: acquire exclusive per-DC lock on write operations:

  • Acquire an exclusive lock on a per-DC basis when performing write operations, both during the creation of a DHCP snippet and its deletion.

  • Always rewrite the DHCP snippet. With the protection of the lock, there is no more need for this check and the library can safely overwrite all the time the DHCP snippet for a given host.

• puppet: add support for puppetserver JSON commands returning non-zero exit code with JSON output (e.g. if a host is missing).

Miscellanea

• doc: add new section for the distributed locking support in the Introduction page.

• doc: mark the module interface as deprecated instead of having the class one as preferred, to better describe the current state.

• tox.ini: remove optimization for tox <4. Tox 4 will not re-use the environments because of the different names, so removing this tox <4 optimization as it's making subsequent runs slower with tox 4+.

• dhcp: simplify tests.

• tests: remove obsolete or not anymore needed items from the false positive list of unused code catched by vulture.

v7.4.1 (2023-10-10)

Minor improvements

• locking: load also ~/.etcdrc for the running user (T341973):

  • We currently save the authentication credential in /root/.etcdrc. Generically load the effective running user's ~/.etcdrc configuration file too and merge it into the one provided in the configuration. This is done best effort, if the ~/.etcdrc file is missing it will be silently ignored.

v7.4.0 (2023-10-09)

New features

• Add distribted locking support (T341973):

  • locking: add new module for distributed locking support via etcd.

  • spicerack: add a new spicerack accessor lock() to get an instance of the locking class to acquire and release cookbook specific custom locks (T341973).

  • cookbook: add --no-locks CLI argument to disable locking acquisition/release on a per-run basis. To be used in case of emergency or if there are issues with etcd that prevents to acquire/release locks properly.

  • By default the locking support is disabled unless the etcd_config is set in the configuration file.

Minor improvements

• spicerack: add owner property to get a pre-formatted string of the form user@host [pid] useful to identify the owner of a current running process.

• spicerack: add current_hostname property to get the hostname of the host where the cookbook is currently running.

• spicerack: improve cookbooks help message:

  • The default argument parser in the CookbookBase class doesn't provide a prog name as it's a bit tricky to guess it because it depends on how many cookbooks are defined in a single file.

  • As a result the help message was not very clear up to now:

    $ sudo cookbook sre.hosts.decommission -h
    usage: cookbook [-h] -t TASK_ID [--force] query
    

  • With this release we inject the cookbook real name in the parser with the additional costruct to use:

    $ sudo cookbook sre.hosts.decommission -h
    usage: cookbook [GLOBAL_ARGS] sre.hosts.decommission [-h] -t TASK_ID [--force] query
    

  • This way it should also help to remind the user that there are global arguments for the cookbook binary in addition to the cookbook-specific ones. It was deemed not necessary to add a message to run cookbook -h to get the available GLOBAL_ARGS, but it can be easily added.

v7.3.1 (2023-10-04)

Bug fixes

• tests: fix test that was actually querying the DNS making it fail in the Debian package build process.

v7.3.0 (2023-10-04)

Minor improvements

• puppet: Add new PuppetServer class and make the PuppetMaster inherit from it as it will be deprecated first and then removed in future releases.

Bug fixes

• decorators: fix the set_tries() function (T346134).

  • It is used to dynamically change the number of tries on a @retry-decorated function/method but was not reading the function signature default value when present. Inspect the signature and if the default value is present, is an integer and is either untyped or typed as integer use it. Add also tests as they were not present and not spotted because the code coverage was considering the function as tested because used in the service module.

Miscellanea

• tests: simplify the spicerack._cookbook.main() tests avoiding to mock the Spicerack instance and using instead the configuration file to instantiate a real instance.

v7.2.2 (2023-09-11)

Minor improvements

• ganeti: add support also for the sandbox VLAN.

• mediawiki: move the calls to noc.wikimedia.org to the kubernetes hosted one.

Bug fixes

• puppet: drop deprecated --ignorecache switch.

• Fix some docstring typos.

Miscellanea

• spicerack: make all CookbookCollection class arguments as keyword-only to avoid mistakes (internal API).

v7.2.1 (2023-06-21)

Bug fixes

• service: make the monitors field of the ServiceLVS class optional to adapt it to the recent change in Puppet about it.

v7.2.0 (2023-05-31)

Minor improvements

• ganeti: add new GanetiRAPI methods nodes() and groups() to get the related info from the cluster.

• ganeti: specify VM memory size in MB to allow for more fine-tune than GB.

• dhcp: when re-generating the DHCP includes and then restarting the DHCP server, in case of a failure make sure to delete the newly created snippet and refresh again to ensure the DHCP is in a good shape.

• dhcp: reword some exception messages.

Miscellanea

• .gitignore: add local config files to it.

• Add Python 3.11 support.

v7.1.0 (2023-05-15)

Minor improvements

• dhcp: expand support for hostname based match using the manufacturer to adapt to different settings.

• remote: improve usability of RemoteHosts.wait_reboot_since() clarifying the message and making it more DRY-RUN friendly.

v7.0.0 (2023-05-08)

API breaking changes

• spicerack: refactor IRC logging:

  • Rename the existing irc_logger to sal_logger as it logs to IRC with the !log and hence to SAL.

  • Add a new irc_logger property to log to IRC on the #wikimedia-operations channel without the !log prefix to just log to IRC and not SAL.

Bug fixes

• doc: do not load UI fix when building the manpage.

v6.4.3 (2023-05-08)

Minor improvements

• ganeti: enable --no-wait-for-sync by default for the virtual machine creation command.

Bug fixes

• decorators: fix dry_run detection that had a bug in the case of a function with a dry_run argument with a default value. The default value was used also in the presence of a an explicit value set by the caller (T335855).

• doc: fix search in documentation as jQuery is not automatically loaded by the rtd theme.

• doc: Remove extra preceding space in intro example.

v6.4.2 (2023-04-17)

Minor improvements

• kafka: remove setting to avoid checking the hostname in TLS certs as all clusters in production are now running with PKI TLS certs that have the hostname in their CN.

Bug fixes

• service: add httpbb_dir field that was added to the Puppet service catalog.

v6.4.1 (2023-03-30)

Bug fixes

• redfish: update log entries location for Dell and make it compatible with different iDRAC versions.

v6.4.0 (2023-03-28)

Minor improvements

• tox: make config compatible with tox 4.x.

• remote: add results to RemoteExecutionError. While waiting for Cumin to support a more robust result reporting, pass the results also in the case of a failed execution to the RemoteExecutionError excepion so that potentially client code could access the partial results on failure using a pattern like:

  try:
      results = remote_hosts.run_sync('some command')
  except RemoteExecutionError as e:
      results = e.results
  

Bug fixes

• setup.py: force dnspython from Bullseye pinning the dependency to the same version of Debian Bullseye as upstream has breaking changes also between minor versions.

• dnsdisc: adapt code and tests to work with dnspython 2.0.0.

• service: improve check_dns_state validation check.

• puppet: make the PuppetMaster class inherit from RemoteHostsAdapter to fix a bug in dry-run mode with a method decorated with @retry.

• service: ensure that dry_run is passed to the Service class to be detected in dry-run mode for methods decorated with @retry.

Miscellanea

• tox: use sphinx-build to generate the documentation, this prevents a deprecation warning for using setup.py.

v6.3.0 (2023-03-15)

New features

• apt: add new module with new AptGetHosts class that inherits from RemoteHostsAdapter to handle simple apt-get use cases but setting all the proper options for non-interactive runs of apt-get.

• spicerack: add new spicerack.apt_get() accessor to run apt-get commands on target hosts.

Minor improvements

• redfish: add simple supermicro class.

• alertmanager: match also FQDN, not only hostnames in the label.

• decorators: add set_tries() function to be used for the dynamic_params_callbacks argument of the @retry decorator to dynamically modify the number of tries to retry from the client.

• dnsdisc: add a resolve_with_client_ip() method to resolve with EDNS Client Subnet (ECS) support.

• service: extend the discovery capabilities of the service catalog to check the DNS records with ECS support adding a check_service_ips() method and a check_dns_state() one.

• spicerack: add authdns_active_hosts property to get a RemoteHosts instance for the authoritative DNS servers currently active. As it uses the Cumin's direct backend it works also if PuppetDB is not available.

Bug fixes

• icinga: handle edge case where status is not optimal but there are no failed services (T330318).

• icinga: uniform code for acked services like failed services to offer the same API in all involved classes.

• k8s: fix existing docstrings.

Miscellanea

• tox: disable bandit's request_without_timeout in tests.

• setup.py: bump dependencies minimum version to match those in Debian bullseye.

• setup.py: remove temporary upper limit for prospector as the upstream issue has been fixed.

• doc: dynamically set copyright year to current year.

• Use GenericAlias objects for type hints in the whole code base given that the lowest supported Python is 3.9:

  • Use directly GenericAlias builtin objects for type hints (e.g. dict[] instead of Dict[]).

  • Use directly GenericAlias objects from the collections.abc module instead of the ones from the typing module (i.e. collections.abc.Sequence instead of typing.Sequence).

  • See also PEP 585.

• docstrings: automatically document type hints using sphinx_autodoc_typehints. Now it's not necessary to repeat in the docstrings the type of the variables and return types as those are automatically added reading the type hints present in the signature. The whole code base has been updated accordingly.

v6.2.2 (2023-02-23)

Bug fixes

• icinga: fix condition that determines if a service status is failed or not (T330318).

• redfish: ensure versions are parsed as packging.version.Version instances.

v6.2.1 (2023-02-20)

Bug fixes

• tests: revert removal of mocked DNS resolver that prevented the tests to run without network access.

v6.2.0 (2023-02-20)

Internal API breaking changes

• spicerack: get authdns servers from config file (T329773):

  • The list of all authdns servers was retrieved via the cumin alias A:dns-auth, which itself comes from Puppet resources (query P{R:Class = profile::dns::auth}).

  • This leads to cookbooks using dnsdisc or service modules failing whenever and authdns is unavailable for maintenance.

  • The source of truth for active authdns servers is hiera, so refactor the modules to use a configuration file populated by Puppet instead.

  • Using the configuration file from Puppet also removes the need to query the IP of the DNS servers and allows to use the Discovery class also withouth a fully working DNS.

  • Use keywords only for most parameters of the touched classes.

  • This change breaks the internal spicerack APIs while the cookbook-facing Spicerack class API has been left untouched.

New features

• alertmanager: add parent Alertmanager class:

  • In some use cases we need to silence alerts in alertmanager that are not attached to any host via the instance label.

  • In order to do so abstract away a higher level Alertmanager class with the generic bits to interact with the Alertmanager APIs and make the existing AlertmanagerHosts class a derived class of that one.

  • Add a new Spicerack accessor alertmanager() to get an instance of a generic Alertmanager without relations to hosts.

Minor improvements

• icinga: allow wait_for_optimal to ignore acknowledged alerts (T319277).

• redfish: allow for refreshing the manager info. Some of the iDRAC info such as firmware and BIOS version are more dynamic and as such we gather them every time, however some other data such as the model is fairly static and can benefit from being cached. As such update the interface so that we can refresh the specific data block for functions that need to.

• redfish: add upload/update methods to push firmware upgrades.

Bug fixes

• mysql_legacy: remove x2 handling logic as it's read-write in both datacenters, and actively written to. Remove it from the module's logic completely to avoid confusion and desync with cumin's list of core-db.

v6.1.0 (2023-02-10)

Minor improvements

• puppet: allow to specify the exact message when disabling/enabling puppet.

• config: expand user's home (~) for logs dir.

• cookbook: improve help message.

• redfish: move Dell specific functionalities to the Dell class.

• redfish: store all OOB info for later use.

• redfish: add system_manager info and properties for bios_version, model, manufacturer.

Bug fixes

• Fix incorrect usage of ClusterShell's NodeSet using the Cumin's nodeset and nodeset_fromlist instead.

Miscellanea

• reposync: switch from copy_tree to copytree.

• kafka: fix typo in docstring.

• dhcp: fix tests using unnecessary hack.

• setup.py: force a newer sphinx_rtd_theme.

• setup.py: pin elasticsearch-curator ~=5.0.

v6.0.0 (2022-12-14)

Configuration breaking changes

• The cookbooks_base_dir config key has been renamed to cookbooks_base_dirs and must be a list of paths.

New features

• Add support for multiple cookbooks paths to be loaded. All the cookbooks paths must have a directory inside named cookbooks/ and this directory must not have an __init__.py file as Namespace Packages are used (see PEP 420) (T325168).

• Add module injection support (T319401):

  • Add an optional configuration key external_modules_dir to define an external modules directory that will be injected in the Python path to allow to use also external modules not present in spicerack.

  • Add a new spicerack.SpicerackExtenderBase class to inherit from in order to define an external accessor class that will be used by Spicerack to allow to use external accessors.

  • Add an optional configuration key extender_class in the instance_params configuration key for specifying the fully qualified name of the Python class to use as the extender class.

Miscellanea

• setup.py: Add python_requires metadata. The latest pyroma does check for its presence and it makes sense to add it to prevent from installing the spicerack package on the wrong Python version.

• setup.py: Revert old upper limit for GitPython, there are no more issue with more recent versions.

• setup.py: Set an upper limit for pylint and prospector for upstream issues.

• setup.py: Split the python auto-formatter test dependencies on their own extra group so that they can be installed alone in the already split virtual environment for the tox envs py3-style and py3-format. This way there are no conflicts between other test dependencies and black and isort.

• setup.py: Add specific style tox environments for each Python version to avoid the CI jobs to pick Python 3.7 that has a pip backtracking issue with the latest versions of the dependencies. Keep the py3-{style,format} environments for ease of use locally and to not break compatibility but make the py3-style one not run automatically in CI.

v5.0.2 (2022-11-17)

Bug fixes

• redfish: fix the reboot message ID check for new iDRAC versions.

v5.0.1 (2022-11-17)

Bug fixes

• redfish: add reboot message ID for new iDRAC versions.

Miscellanea

setup.py: remove support from Python 3.7 and 3.8. tox: remove support from Python 3.7 and 3.8.

v5.0.0 (2022-11-10)

Dependencies breaking changes

• Starting with Spicerack v5.0.0 the support for Python 3.7 and 3.8 is dropped. For now there are no breaking changes but it's not guaranteed to work with those versions anymore.

API breaking changes

• constants: remove CORE_DATACENTERS constant:

  • Remove the constant from Spicerack as it's a duplicate of the one already present in wmflib.

  • Convert all Spicerack code to use the same variable from wmflib.

  • All the cookbooks have been already migrated to use the wmflib one.

Minor improvements

• ipmi: clarify that the target can also be an IP address. The ipmi module works the same as with a management FQDN.

Bug fixes

• netbox: update allowed state transitions:

  • As the way we use Netbox status is changed as part of the work in T320696 and the staged status is not anymore used, update the allowed transitions based on the new Server Lifecycle Diagram.

Miscellanea

• mypy: remove upper limit and refactor mypy configuration to properly work with newer versions.

v4.0.0 (2022-09-28)

API breaking changes

• redfish: use the management IP instead of FQDN to connect to the management console:

  • Some DELL hosts come with the idrac.webserver.HostHeaderCheck setting set to 1, that prevents to connect to the Redfish API unless the hostname is set in the configuration, creating a chicken and egg problem to automate the initial setup of the hosts.

  • To prevent this switch the whole module to use directly IPs for now. We might want to improve this later setting the hostname in the iDRAC settings and then switching to use the FQDN once that is configured, but because most of the automation will be already done by that time it's not clear if it would be a real win.

  • [BREAKING API] this changes the spicerack.Spicerack.redfish() signature to require a hostname instead of a management FQDN and also makes the username parameter optional, defaulting to use root.

  • [BREAKING API] this changes the spicerack.redfish.Redfish class signature to require a hostname and management IP address instead of a single parameter with the FQDN. Although breaking, no cookbook usage should instantiate this class directly, but always via the above accessor.

Minor improvements

• icinga: add explicit support of the DRY-RUN mode (T315537):

  • While the DRY-RUN compatibility of the icinga module was guaranteed by the remote module, there was a usage of the @retry decorator that wasn't able to detect when in DRY-RUN mode and accordingly reduce the number of retries.

• Bump pynetbox dependency to ~= 6.6 (T310745).

• netbox: enable pynetbox threading (T311486).

Miscellanea

• doc: fix sphinx_checker script for Python 3.10.

• doc: add an example on how to use the TOX_SKIP_ENV environmental variable to run only certain tox environments when in development.

• doc: improve documentation of the CookbookBase classes usage.

v3.2.1 (2022-08-31)

Bug fixes

• elasticsearch_cluster: simplify routine to start masters last. Due to the multiple clusters an host can be a master in one instance and a child of another instance, bringing the process to a halt using the previous logic. The new logic returns all the hosts that are child for all instances first and after that the remaining ones that are master for at least one instance.

• peeringdb: minor fixes:

  • Make the Spicerack.peeringdb() accessor more flexible allowing the configuration file to miss non mandatory keys.

  • Add tests for the Spicerack.peeringdb() accessor.

  • Use empty string as default value for the token to avoid the Optional type.

  • Fix mypy ignore for type mismatch.

  • Fix various docstrings.

Miscellanea

• CHANGELOG: fix typos and uniform format.

v3.2.0 (2022-08-18)

New features

• peeringdb: add a new module to interact with the PeeringDB API.

Minor improvements

• elasticsearch_cluster: ensure to restart masters one at a time.

Miscellanea

• flake8: move flake8's configuration all into setup.cfg.

v3.1.1 (2022-07-26)

Bug fixes

• k8s: Increase retry value to prevent timeouts.

Miscellanea

• Add support for python 3.10.

v3.1.0 (2022-07-20)

Minor improvements

• redfish: add support to check the reboot of the DELL iDRACs:

  • add a most_recent_member() method in the Redfish class to return the most recent message from an API reply with members from Dell.

  • add a last_reboot() method to the Redfish class to get the time of the last DELL iDRAC reboot.

  • add a wait_reboot_since() method to the Redfish class to poll until the DELL iDRAC comes back online after a reboot.

• redfish: add property for the HttpPushURI url, needed for pushing firmware to the DELL iDRACs.

• redfish: add a generation property to the Redfish class to represent the DELL iDRAC genration i.e. 13 == idrac8, 14 == idrac9, and allow us to implment workarounds for older generations.

• redfish: add a fqdn() getter property and __str__() method to the Redfish class:

  • When passing around a Redfish instance it's useful to know what host it represents as such add a getter for the FQDN property and update the __str__() metbod to also return the FQDN.

• k8s: Add KubernetesNode.taints propertry to return the taints of a node.

• k8s: Retry checks for expected pods on drain as in some cases (e.g. pods not catching TERM) it might take a while for pods to actually terminate. Retry the check for expeced pods to reduce the chance for errors.

• k8s: Retry pod evictions on HTTP 429 from API server:

  • An HTTP 429 response from the API server means that the eviction is not currently allowed because of a configured PodDisruptionBudget or a API server rate limit was hit. Retry evict() calls in both cases 3 times with exponential backoff.

• tests: reduce runtime by more than 80%:

  • The logging module setup performed in the spicerack._log.setup_logging() function is not automatically reset by pytest, leading to slowness in some tests, in particular those with a lot of output, for example due to a lot of retries.

  • Add a _reset_logging_module() funtion in the tests for the _log module that removes all exisiting filters and handlers to both the root and the IRC loggers.

  • Call the _reset_logging_module() function in the teardown of every test that directly or indirectly calls the spicerack._log.setup_logging() function.

  • This reduces the runtime of the unit tests by more than 80%, in my local environment for example it went from ~150s to ~25s for the 825 tests run.

Bug fixes

• redfish: better compare Dell SCP attributes:

  • When comparing Dell SCP attributes for the configuration, consider them identical if they are a comma-separated list both if the separator is just the comma or comma+space. Some versions of iDRAC return the values comma+space separated when getting the current configuration.

• tests: fix caplog usage:

  • Make sure to use caplog.at_level() every time the pytest caplog fixture is used to ensure the reliability of the test itself and to avoid altering the level for other tests.

  • Rename the argparse.py test cookbook to argparse_ok to prevent any conflict with the stdlib argparse module.

v3.0.0 (2022-06-28)

API breaking changes

• ganeti: refactor the Ganeti module to support the new data model in Netbox:

  • With the new representation of Ganeti data in Netbox, the hardcoded matching between cluster names and Ganeti RAPI FQDN endpoint would not work anymore.

  • Refactor the module to gather the data directly from Netbox.

  • This requires the addition of a custom field ip_address for the virtualization cluster groups model that connects it to the Ganeti RAPI VIP "svc" DNS name that is assigned to the related IP address in Netbox. The custom field has been already added and populated in Netbox in production.

  • The main benefit is the removal of the hardcoded mapping between clusters and their groups (rows/racks).

  • Add a new get_cluster() and get_group() methods in the Ganeti class to get a new GanetiCluster or GanetiGroup dataclass instances that represent the data required to identify the related resources.

  • Removed the hardcoded magic logic that mapped a row A to a Ganeti group row_A as we're moving away from row-level redundancy at the network layer towards a rack-level redundancy model. This allows to rename the Ganeti groups at anytime freely.

Minor improvements

• icinga: ensure that the downtime was applied (T309447):

  • Add a wait_for_downtimed() method that polls the Icinga status to ensure that the hosts got downtimed.

  • Do this best effort, just logging a warning for now in case the downtime can't be verified.

Bug fixes

• redfish: make task polling work with older models that set the end time to Unix epoch at the task start.

Miscellanea

• log: stop suppressing logging exceptions, that were silenced in the logging configuration.

• doc: fix intersphinx links.

v2.6.0 (2022-06-07)

Minor improvements

• redfish: Assume all GET and HEAD requests are read-only and anything else is potentally read-write.

• redfish: allow to submit tasks with DELETE as some Redfish REST API DELETE actions do submit jobs. The submit_task() method accepts an HTTP method different than POST now.

• netbox: update netbox to use internal discovery address as it got migrated from a public IP to the discovery infrastructure.

Miscellanea

• doc: set default language as Sphinx 5.0+ requires language to not be None when warnings are treated as errors.

• pylint: remove unnecessary comments. The latest pylint has moved the no-self-use reported issue to an optional plugin. We don't need to enable it, hence removing the unnecessary comments.

v2.5.0 (2022-05-26)

API breaking changes

• redfish: update signature of the request() method to support dynamic keyword arguments that will be passed directly to the requests library:

  • Although this breaks backward compatibility of the existing API for the request() method, it's not currently used directly anywhere and so it was deemed ok to not justify a new major release for this.

  • In particular the previous data parameter that was passed to requests's json parameter would now be passed to request's data parameter, so not being automatically converted to JSON. Existing calls have been modified to call requests() with a json parameter instead.

New features

• service: add new module to expose Puppet's service::catalog:

  • Add a new module to load the Puppet service::catalog hieradata structure into Spicerack.

  • Part of the abstractions allow to access in a more programmatic way the properties of a given service.

  • It also allow to depool/pool (and related context manager) a service in the DNS Discovery realm.

  • It also allow to downtime (and related context manager) a service in a given datacenter in Alertmanager.

  • See the service module example usage.

Minor improvements

• reposync: improve git push error handling catching more possible git errors.

• ganeti: add a startup() method to startup a Ganeti VM (T306661).

• ganeti: add set_boot_media() method to modify the instance boot media and change it between disk and network (PXE) (T306661).

• ganeti: print the output of a Ganeti VM creation while it's being created so that it gets printed live and not at only at end.

• dhcp: add to the DHCPConfOpt82 and DHCPConfMac classes a media_type parameter:

  • This new media_type parameter will allow use to easily choose PXE boot media other then the default debian installers. Specifically this will allow us to create cookbooks to test specific point releases as well as rescue and secure-wipe options.

Bug fixes

• mediawiki: Mediawiki APIs now are only listening only on HTTPS, call the siteinfo API in HTTPS.

• remote: increase the wait for reboot timeout (T307260):

  • In some cases, in particular during reimages, the reboot time can take longer. Increase the limit for now as in most cases this will not change anything as the check will succeed way before the timeout.

Miscellanea

• tests: fix yaml file indentation.

• doc: fix typo.

• setup.py: mark the module as typed so that mypy can type check calls in other tools that are importing this library.

v2.4.1 (2022-04-12)

Minor improvements

• elasticsearch_cluster: don't wait for green on first node.

• alertmanager: improve downtime:

  • Allow to pass hosts with already a specific port. If the port is present no port-related regex is added, if the port is not present the port-related regex will be automatically added.

  • Optimize the regex adding just once the port regex at the end if all hosts don't have the port specified.

  • Add a matchers parameter to the downtime() and downtimed() methods to allow to perform additional filtering adding additional matchers.

  • Raise an error in case an additional matcher is trying to target the instance property.

Bug fixes

• alertmanager: fix downtime:

  • Fix the way the matchers for the silence are created. Because AlertManager and Prometheus will evaluate all matchers in AND, we can only add one single matcher for the instance property, that has to match all given hosts, as opposed to the current implementation that was adding one matcher per host.

v2.4.0 (2022-04-04)

New features

• k8s: add a new module with initial support for Kubernetes that supports draining a node (T300879).

• spicerack: add a new Spicerack.thanos() accessor to get an instance of wmflib.prometheus.Thanos.

• ipmi: add a remove_boot_override() method to clear any BIOS boot parameter override because some hosts don't automatically clear that after a reboot.

Minor improvements

• ipmi: improve the force_pxe() method changing the way it sets the Force PXE bit in the BIOS boot parameters to force the reset of the valid flag after a reboot and consider the valid flag as harmless anyway (T304434).

Miscellanea

• pylint: fix newly reported issue.

v2.3.3 (2022-03-17)

Minor improvements

• reposync: don't catch the RepoSyncNoChangeError allowing the calling cookbook to decide what to do in case of no changes in the repository.

• reposync: add a force_sync() method to perform a force push from the local repository to all remotes.

v2.3.2 (2022-03-10)

Bug fixes

• alertmanager: add missing support for dry-run mode.

• reposync: make tests run quicker:

  • Some tests were using 192.0.2.1 as a git remote, that doesn't fail immediately, at least on macOS. Replace it with a non-existent local path.

v2.3.1 (2022-03-10)

Minor improvements

• spicerack: make http_session more flexible:

  • Instead of updating the signature with the new parameters available in wmflib, relax the signature here in spicerack and delegate to wmflib what are the accepted parameters.

Bug fixes

• alertmanager: do not retry on HTTP 500 responses:

  • The Alertmanager API can respond with an HTTP Status Code of 500 on some requests with a valid JSON response, although there was no server error (i.e. trying to delete an already deleted silence).

  • Do not retry on 500 responses, allowing requests to get a proper response and then let the module itself decide what to do based on the content of the response.

v2.3.0 (2022-03-09)

Minor improvements

• alertmanager: catch the already deleted silence error (T293209):

  • The Alertmanager API, when trying to delete an existing silence, returns 500 with a JSON string message in the case of an already expired or deleted silence.

  • On delete, catch the exception and just log a warning message in case the silence has been already deleted / is already expired.

  • In orther to achieve this, change the AlertmanagerError exception to accept an optional parameter with the API response object.

• elasticsearch_cluster: load the configuration from a yaml file, remove the hardcoded one (T278378).

Miscellanea

• spicerack: use the private property for the config dir within the class, for coherence.

v2.2.0 (2022-03-08)

New features

• alertmanager: introduced a new module to manage resources on AlertManager (T293209):

  • It has an AlertmanagerHosts class that currently supports creating a silence (downtime in Icinga terminology) and removing it given its ID. It also provides a context manager to perform the silence similarly to the icinga module.

• alerting: introduced new alerting module with an AlertingHosts class as a wrapper around the IcingaHosts and AlertmanagerHosts classes so that the same actions are performed on both instances.

• spicerack: add accessors for the new AlertmanagerHosts and AlertingHosts classes as alertmanager_hosts and alerting_hosts respectively. The preferred way is to use the alerting_hosts accessor so that actions like the downtime are performed on both systems.

Bug fixes

• redfish: fix the default value for the allow_new_attributes parameter of RedfishDell.scp_dump().

v2.1.0 (2022-03-03)

New features

• reposync: add new module to manage syncing of automatically generated repositories.

Minor improvements

• redfish: DellSCP, allow creation of new entities:

  • So far the DellSCP class allowed only to modify existing attributes in existing components.

  • When dealing with a DellSCP configuration, there are cases in which it might be necessary to create attributes that do not exist in the current configuration. For example when changing the boot mode between Bios and Uefi a long list of attributes disappear/appear in the configuration.

  • To allow this use case an allow_new_attributes keyword only parameter has been added to the constructor to explicitly allow new attributes, keeping the existing behaviour of typo-protection if that is not passed.

  • Another possible use case is to start from a configuration and create a components section from scratch.

  • To allow this use case an empty_components() method was added that, while keeping the rest of the configuration intact, empties the existing components and from there allows to set new attributes, transparently creating any missing component.

  • Add the allow_new_attributes parameter to RedfishDell.scp_dump() to enable this new feature when dumping a configuration.

Bug fixes

• dhcp: fix lowercase serial tag matching.

Miscellanea

• setup.py: temporary limit redis library:

  • The latest redis release v4.1.4 creates some dependency issue, for now limit the upper version as we're anyway using v3 in production as that's the version up to Debian Bullseye.

• setup.py: upper limit for black:

  • On Debian bullseye elastcisearch-curator latest release dependencies have a conflict with black's dependencies and it's not possible to put an upper limit to elastcisearch-curator because previous version don't build properly on Bullseye from pip (the debian package version of it has a patch to override its dependency constraints).

  • To prevent conflicts force an upper limit on the black version for now.

• bandit: ignore hardcoded password in tests:

  • Ignore the B105:hardcoded_password_string and B106:hardcoded_password_funcarg checks in test directories.

  • Removed related #nosec comments unnecessary now.

• prospector: ignore deprecation message:

  • The latest prospector issues a deprecated message for the pep8 and pep257 tools that have been renamed to pycodestyle and pydocstyle respectively. The new names are incompatible with prospector < 1.7.0, so for now keep the old names and disable the deprecation warning.

v2.0.0 (2022-02-15)

API breaking changes

• management: removed module, it was deprecated in v1.0.0.

New features

• spicerack: allow to execute another cookbook from within a cookbook:

  • Add the capability from within a cookbook to call another cookbook with custom parameters using the run_cookbook() method in the Spicerack class.

  • The called cookbook will be executed with the same global options with which the current cookbook is running with and will log in the same file of the current cookbook run.

Minor improvements

• redfish: better support of parsing JSON responses (T299123):

  • In some older Dell servers the Redfish API sometimes replies with different casing for the MessageId key, like MessageID.

  • It's also possible that Oem custom messages are reported in the same replies with a different structure.

  • Skip the Oem messages and try both keys cases when parsing the reply.

• redfish: improve support for DRY-RUN mode:

  • In DRY-RUN mode allow read-only requests to be performed (only GET and HEAD) but return a dummy successful responses in case of an exception raised by requests (timeout, connection error, etc).

  • In DRY-RUN mode don't allow read-write requests and return a successful dummy response instead.

  • In various methods return a dummy response in DRY-RUN mode.

• dhcp: case-insensitive match of the serial number for the Dell management DHCP requests:

• When matching the serial number in the DHCP request for the management interfaces of Dell servers, match them in a case-insensitive way because the data sent varies between hosts (idrac-ABC1234 or iDRAC-ABC1234).

Miscellanea

• setup.py: the latest v2.2.0 release of dnspython is generating mypy issues, temporarily put an upper limit to it.

• spicerack: adapt type hint to the latest wmflib release.

v1.1.1 (2021-12-22)

Minor improvements

• redfish: tell if any change was made in DellSCP instances:

  • When updating a DellSCP configuration with the set() or update() method, return True if the config was actually changed, False if it had already the correct value(s).

Bug fixes

• dhcp: fix file removal check in dry-run mode.

v1.1.0 (2021-12-16)

New features

• spicerack.redfish: add new module with support for Redfish API:

  • Add a new redfish module that allows to interact with the Redfish API. As Redfish implementation differs sensibly between vendors, there are some basic functionalities in the Redfish class and then there is a RedfishDell class for Dell-specific functionalities.

  • At the moment the only supported vendor is Dell (hence the hardcoded RedfishDell call in Spicerack.redfish().

• spicerack: add a management_password property getter to access the cached management password. If the cache is empty the password will be asked to the user.

Minor improvements

• ganeti: add new Ganeti clusters in the new site drmrs.

Bug fixes

• ipmi: when running an IPMI command that contains sensitive data, allow to hide the sensitive data from the logs and the outputs.

• ganeti: fix up row configuration for ganeti test cluster.

• dhcp: fix missing semicolon in DHCP config.

• remote: intercept bad uptimes in wait_reboot_since().

  • In some cases the uptime method could fail to parse the host uptime, for example during a shutdown of a system where the login might be prevented to the host.

  • Make sure that the wait_reboot_since() method catches those errors too and retries.

Miscellanea

• Adopt pathlib.Path instead of the os and os.path functions across the project to modernize it following current best practices.

• administrative: add examples to the documentation and documentation for the special method __str__.

• pylint: fix newly reported issues.

v1.0.6 (2021-10-21)

New features

• dhcp: add support for MAC address based config (T269855):

  • Add support for MAC address based configuration snippets to be used in the automation for Ganeti VMs instead of using DHCP Option 82 as the MAC address is retrieved from Ganeti API.

  • The MAC address is validated to ensure has the format accepted by the DHCP server.

  • Consolidate the filename path for both DHCP Option 82 and MAC address based configuration to be in the same directory, dependent only by the TTY settings as there is no other difference between the two and it allows to prevent duplicated snippets for the same hostname in different directories as the library checks that the file doesn't exists before creating it.

  • Consolidate the defult string representation implementation of the DHCPConfiguration derived classes into the abstract parent one because they are all the same. Define a class property _template as part of the DHCPConfiguration class API.

Minor improvements

• mediawiki: add a get_primary_dc() method that returns the primary/active datacenter.

• kafka: docstrings minor improvements.

Miscellanea

• changelog: fix typo in previous entry.

v1.0.5 (2021-10-12)

New features

• kafka: add a new kafka module with the following capabilities (T291681):

  • transferring of offsets between consumer groups and clusters approximating offsets based on timestamp.

  • approximating and seeking offsets based on user provided timestamps.

Minor improvements

• icinga: add recheck_failed_services() method to force a recheck of services which are in failed state.

Bug fixes

• puppet: get only the last line of output in PuppetHosts.get_ca_servers() to ignore spurious output that might be present in some environments.

v1.0.4 (2021-10-06)

New features

• dhcp: use IP address instead of DNS name:

  • Given that all the required data comes from Netbox there is no point to depend on the DNS when generating the DHCP snippets, require to pass the IPv4 instead of the FQDN.

  • Renamed fqdn parameter to ipv4 in the DHCPConfOpt82 class.

  • Renamed ip_address parameter to ipv4 in the DHCPConfMgmt class.

  • Although technically this is an API change, the whole module is new and still unused except from the experimental reimage cookbook, hence not considering it as a breaking change for the semantic versioning.

Minor improvements

• remote: reduce wait time for reboot to 20 minutes.

v1.0.3 (2021-09-28)

Bug fixes

• dhcp: fix typo in opt82 file path.

v1.0.2 (2021-09-27)

Minor improvements

• dhcp: always require to se the OS version when instantiating a DHCPConfOpt82 instance. Although technically this is an API change, the whole module is new and still unused, hence not considering it as a breaking change.

• remote, puppet: reduce logging verbosity.

Bug fixes

• ganeti: use --force option in shutdown method when calling gnt-instance shutdown to work with all states a VM can be in.

• puppet: fix check exception inheritance to the correct SpicerackCheckError.

v1.0.1 (2021-09-23)

Minor improvements

• remote: refactor wait_reboot_since():

  • As the check for uptime is currently either returning a value for all hosts or raising an exception, remove the existing logic to check for a partial result as that can't happen.

  • Catch instead the error and re-raise a check exception with a clear message.

  • Also round the printed value of the uptime and the time against which it's checked to 2 decimal values for more readability.

Miscellanea

• setup.py: limit elasticsearch max version:

  • The latest 7.15.0 release has started to deprecate things for the upcoming 8.0.0 release, and mypy started complaining about some return types.

  • Instead of fixing the signatures to be compatible with both versions put a max version limit for now, we'll deal with the upgrade when the time will come, Debian most recent version is 7.1.0.

v1.0.0 (2021-09-22)

API breaking changes

• remote: remove RemoteHosts.init_system() method:

  • As systemd is used by all hosts and this method is not used in any cookbook, remove it completely as it's no longer needed.

New features

• remote: add support to enable/disable Cumin output:

  • Add support to suppress Cumin's output and progress bars independently to the RemoteHosts and LBRemoteCluster classes.

  • Add a print_output and print_progress_bars boolean parameters to run_sync(), run_async() and run() methods to independently print Cumin's output and progress bars respectively.

  • Add a simplified verbose parameter to the more higher level methods restart_services() and reload_services() that when set to False will suppress both output and progress bars at once.

  • Add just the print_progress_bars parameter for the high level methods wait_reboot_since() and uptime().

  • All the new parameters default to True right now to keep the existing behaviour, to be changed to False in a future release.

Minor improvements

• icinga: reduce verbosity of Cumin's output, taking advantage of the new parameters to control the output of Cumin's commands.

• puppet: reduce verbosity of Cumin's output, taking advantage of the new parameters to control the output of Cumin's commands.

• dhcp: reduce verbosity of Cumin's output, taking advantage of the new parameters to control the output of Cumin's commands.

Bug fixes

• ipmi: improve dry-run mode for force_pxe():

  • When force_pxe() can't verify that the next boot will indeed be via PXE it raises an exception. Convert that into a warning logging message when in DRY-RUN mode to let the cookbooks continue the DRY-RUN.

Miscellanea

• versioning: moving Spicerack releases to a semantic versioning schema.

• management: deprecate the Management class:

  • As its only purpose was to get the management FQDN of a host, given that the same functionality is now provided by the netbox module via the NetboxServer class and its mgmt_fqdn and asset_tag_fqdn properties, deprecate the class for a subsequent removal.

• confctl: fix example code in docstring.

• pylint: fix newly reported issues.

• doc: add how to contribute section.

v0.0.59 (2021-09-09)

API breaking changes

• ipmi: refactor class signature:

  • API breaking change, but the Spicerack.ipmi() accessor is used only in the sre.hosts.decommission and sre.hosts.ipmi-password-reset cookbooks, so it should be trivial to change both at once.

  • Convert the IPMI class to require the FQDN of the management console to target, to avoid the need to pass that around both from the client and internally in the class.

  • The caching of the management password is done transparently by the Spicerack.ipmi() accessor to avoid the anoyance of being asked the management password for each host.

• dhcp: small refactor (the module is still unused):

  • Rename switch_port to switch_iface to avoid confusions.

  • Rename the context manager from dhcp_push() to config() as it's more natural to use: with dhcp.config(my_config): # do something.

  • Simplify formatting of templates, added ignores to vulture for false positives

  • Add constructor documentation to the dataclasses.

• icinga: remove the deprecated Icinga class:

  • The Icinga class has been deprecated for a while now and it's time to remove it completely. No cookbook is using it anymore.

New features

• remote: add support for the installer key:

  • When instantiating a remote() instance, allow to pass a new parameter installer, defaulted to False, that when True will use the special installer key for the remote instances that allow to connect to the Debian installer environment or a freshly installed host prior to its first Puppet run.

• ipmi: add status and reboot capabilities:

  • Add a new method power_status() that returns the current power status and is also used by the existing check_connection() method.

  • Add a new method reboot() to issue an IPMI power on or power cycle, based on the current status of the device.

• netbox: add getter asset_tag_fqdn for the asset tag mgmt FQDN property.

• icinga: add downtime_services() and remove_service_downtimes() and also a services_downtimed() context manager to allow to downtime only the host services that matches the given regex.

Minor improvements

• puppet: minor improvements:

  • Return the results from the Puppet.first_run() method to allow to save it to a file like the current reimage script does.

  • Add an accessor for the master_host property in the PuppetMaster class as this is created and instantiated by Spicerack and was hidden from the user of the API.

• decorators: migrate to the wmflib version of @retry (T257905):

  • Use the wmflib version of @retry while keeping the dry-run awareness and default to catching SpicerackError instead of WmflibError like the pre-exsiting version was doing.

Miscellanea

• code style: migrate all the usage of string format() to f-strings.

• pylint: addressed newly reported pylint issues and removed unnecessary disable comments.

• prospector: disable E203 for pep-8 over black.

• code style: if there are no local modifications check last commit instead of not checking anything.

v0.0.58 (2021-08-25)

New features

• Class API: add rollback() method

  • Add a new rollback() method to the CookbookRunnerBase base class that by default does nothing.

  • The method is called by Spicerack when a cookbook exits with a non-zero exit code or raises an un-caught exception.

  • This allows cookbooks to define their own cleanup strategy in case of errors, for example to restore a previously coherent state.

  • Any exception raised by the rollback() method will be caught and logged by Spicerack with its original exit code and will then exit with a reserved exit code for a failed rollback.

Minor improvements

• mediawiki: remove cron-specific maintenance implementation details, replaced by systemd timers (T289078).

Bug fixes

• icinga: use shlex to quote the command string for bash (T288558):

  • This fixes the downtiming that would fail if the admin reason contains an apostrophe, due to lack of escaping.

• mediawiki: ignore php-fpm when stopping cronjobs (T285804):

  • On mwmaint, php-fpm is used to serve noc.wikimedia.org so we want to keep it running even when stopping cronjobs.

v0.0.57 (2021-08-02)

Minor improvements

• dnsdisc: improved message logged explicitely saying what was checked and what didn't match when checking that a discovery record has been updated (T285706).

• icinga: adapt to the newer API of the icinga-status output.

• icinga: write directly to the Icinga command file instead of calling the icinga-downtime wrapper script where it was used so that the whole module now interacts directly with the Icinga command file. This opens up the route for further improvements (T285803).

• ganeti: add ganeti test cluster to the possible Ganeti locations (T286206).

• mysql_legacy: re-add x2 database section and add support for active/active core sections (T285519):

  • get_core_dbs() now supports excluding sections from its cumin query. All of the functions that call it in the context of setting the database read-only or read-write will now exclude sections listed in ACTIVE_ACTIVE_SECTIONS.

Bug fixes

• puppet: when regenerating the client certificate, do not rely on the exit code of the Puppet command as it might be misleading. It already relies on successfully finding the certificate fingerprint.

Miscellanea

• tox: remove flake8-import-order plugin as dependency now that the import order is ensured by black and isort.

v0.0.56 (2021-06-26)

Bug fixes

• mediawiki: reverted the change of v0.0.55 to make siteinfo API request over HTTPS.

• mediawiki: remove unnecessary and broken disable of systemd timers added in version v0.0.55.

• mysql_legacy: reverted the change of v0.0.49 to add the new x2 database core section (T285519).

v0.0.55 (2021-06-24)

API breaking changes

• mediawiki: Update cronjob code now that most are systemd timers:

  • Removed check_cronjobs_enabled().

  • Renamed stop_cronjobs() to stop_periodic_jobs().

  • Added check_periodic_jobs_disabled(), check_periodic_jobs_enabled() and check_systemd_timers_enabled().

Bug fixes

• mediawiki: Make siteinfo API request over HTTPS.

v0.0.54 (2021-06-21)

API breaking changes

• icinga: rename some IcingaHosts methods:

  • This is an API breaking change, but the newly introduced IcingaHosts API is not yet used widely, just one Cookbook uses it so far.

  • Rename some methods of the IcingaHosts class to be more dry and explicit. Namely: * hosts_downtimed -> downtimed (context manager) * downtime_hosts -> downtime * host_command -> run_icinga_command

v0.0.53 (2021-06-10)

Bug fixes

• icinga: use bash wrapper to allow sudo in the IcingaHosts class.

Miscellanea

• doc: use add_css_file() instead of add_stylesheet().

• doc: fix parameter type in docstring.

v0.0.52 (2021-05-06)

New features

• dhcp: Add module for manipulating dynamic DHCP entries on target data centers and restarting the DHCP server (T269855).

• icinga: pass verbatim_hosts option to the icinga-status script when using verbatim Icinga hostnames that are not real hosts.

Bug fixes

• netbox: fix check for server role:

• The physical devices and virtual machines objects in Netbox have different names for the role property (device_role vs role). Use the correct property each time.

• icinga: fix typo in docstring.

v0.0.51 (2021-05-04)

Minor improvements

• dnsdisc: do not configure DNS resolver. As the module is injecting the nameservers of the authoritative DNS, do not let the DNS module auto-configure itself with /etc/resolv.conf.

Bug fixes

• tests: fix mock of the DNS module that was not in some cases properly mocked and the tests were relying on a properly configured /etc/resolv.conf.

v0.0.50 (2021-05-04)

Dependencies breaking changes

• setup.py: relax elasticsearch dependencies:

  • In order to be able to build spicerack for Debian bullseye that ships python3-elasticsearch 7.1.0 and python3-elasticsearch-curator 5.8.1, relax the related dependency constraints in setup.py.

  • Elasticsearch requires to bump the version above the suggested compatibility matrix, we'll test if all works as expected. See the elasticsearch compatibility matrix.

  • Elasticsearch curator matches upstream compatibility matrix, see the elasticsearch curator compatibility matrix.

  • As Spicerack is released via debian packages this will not affect the buster builds.

API breaking changes

• netbox: improve as_dict():

  • Instead of calling serialize() for the conversion to dictionary, just calling dict() on the object gives a more useful representation of the object because all the nested properties are converted to string or sub-dictionaries with useful values instead of just the IDs.

  • As a result any usage of as_dict() that relied on the format of specific fields might break. At the moment no cookbook is using it.

  • See also the "Casting the object as a dictionary" example in pynetbox.core.response.Record.

New features

• netbox: add NetboxServer class:

  • Add a NetboxServer class in the netbox module to give a higher level abstraction across physical servers and virtual machines.

  • This is particularly useful to finally have an authoritative way to convert a hostname into a FQDN or get the managment FQDN of a host given its hostname (T240176).

  • The class also allow to update the device status only if it's a physical host and the status transition is approved.

  • Those new features will be used by the cookbook that will replace the reimage script and then the current usage of some of the existing methods in the Netbox class should be converted to use this class instead.

• icinga: add new IcingaHosts class (T277740):

  • Implements the TODO that wanted to move the Icinga class into a class that is initialized with the target hosts so that it's not necessary anymore to pass them to each method.

  • Keep the existing Icinga class for now, but mark it as deprecated, both in the documentation of spicerack.Spicerack.icinga() and icinga.Icinga() and emit also a DeprecationWarning when instantiated. It will be removed in the next release once all the cookbooks have been migrated to the new spicerack.Spicerack.icinga_hosts() accessor.

  • Move the detection of the Icinga command file to its own class to allow to cache it across different instances, making the instantiation of multiple IcingaHosts class free after the first one.

  • Allow to manage also non-servers that are defined as Icinga hosts passing the verbatim_hosts parameter, that will not extract the hostname from the given hosts assuming that they are already FQDNs.

• toolforge.etcdctl: Allow getting the cluster health. This opens up being able to wait/stop if the cluster status is not what's expected when doing operations (T276338).

Minor improvements

• icinga: use a bash command wrapper to allow sudo, otherwise the echo command will fail to output to the file.

• icinga: use a sudo-friendly command to detect the Icinga command_file.

• netbox: improve as_dict():

  • Instead of calling serialize() for the conversion to dictionary, just calling dict() on the object gives a more useful representation of the object because all the nested properties are converted to string or sub-dictionaries with useful values instead of just the IDs.

  • See also the "Casting the object as a dictionary" example in pynetbox.core.response.Record.

Bug fixes

• remote: fix use_sudo on split().

• netbox: fix object type returned for status. The status should be returned as string and not as a Netbox object.

• doc: add documentation for the toolforge package.

• doc: remove obsolete configuration.

• setup.py: add missing tag for Python 3.9, already supported.

• tests: fix pip backtracking separating the prospector tests into its own virtualenv.

• tests: fix format checking:

  • If no Python files were modified at all, the latest isort would bail out. Skipping the checks if no Python files were modified at all.

• doc: fix documentation checker for sub-packages:

  • The existing checker was assuming a flat space of modules inside spicerack, while now we have also subpackages. Adapt the checker to detect those too.

  • Convert file operations to pathlib.

Miscellanea

• doc: move ClusterShell URL to HTTPS.

• netbox: refactor unit tests.

v0.0.49 (2021-03-04)

API breaking changes

• icinga: changed the type for the hosts parameter in the get_status() method from spicerack.typing.TypeHosts to cumin.NodeSet.

New features

• icinga: add Icinga.wait_for_optimal() method to pause while hosts converge to an optimal state.

• puppet: add Puppet.get_ca_servers() method to retrieve the configured Puppet ca_server on the target hosts.

• remote: allow prepending every command to execute on the target hosts with sudo. This is a first temporary iteration until Cumin will support it natively.

• toolforge.etcdctl: add new toolforge package with an etcdctl module to run etcdctl commands and retrieve a parsed output. Focused on etcd member management only for now (T267412).

Minor improvements

• config: allow to use paths relative to the user's $HOME directory expanding ~.

• logging: improve logging format:

  • Add the DRY-RUN prefix also to file logs to allow to distinguish dry-run executions from the real ones just looking at the logs.

  • Improve the execute cookbook log message including the whole arguments so that it includes also the global args such as verbose and dry-run.

• remote: RemoteHosts.wait_reboot_since() is now using a constant backoff. Previously, a linear backoff with a base delay of 10 seconds was used. Since we do expect the reboot of a server to take some time, by the time the server has rebooted, the retry interval has already grown to multiple minutes. A constant backoff should be appropriate and should increase the reactivity of this check significantly.

• mysql_legacy.py: Add the new x2 database core section (T269324).

Bug fixes

• cookbooks: force the title to be one line. When reading the title from the cookbooks, pick only the first line to prevent the UI to be cluttered by a title erroneously set to multi-line.

• tox: fix for when the system setuptools is too old.

• elasticsearch_cluster: Revert the return the cluster name in ElasticsearchCluster.__str__ change added in v0.0.32.

• remote: fix pylint typing confusion.

Miscellanea

• gitignore: add vim swap files.

• tests: temporary force mypy upper version to avoid a regression in release 0.800.

• tests: tox, enable python 3.9 support.

• code style: introduced black and isort as autoformatters (T211750).

• doc: add a development page to highlight how the code is formatted and how to integrate the code formatters with an editor/IDE or in the git workflow (T211750).

• git: allow exclude code auto formatters refactor commit from git blame adding the .git-blame-ignore-revs file.

v0.0.48 (2021-01-18)

Bug fixes

• logging: fix base path and name to setup logging.

  • In the recent refactor to the new APIs, the paths passed to the setup_logging function were not anymore correct. Now that the cookbook items have a proper Spicerack-formatted path and name, use them directly.

v0.0.47 (2021-01-13)

API breaking changes

• Use newly migrated code from wmflib:

  • Some additional functionalites were moved to wmflib (>= 0.0.5), remove the duplicated code from Spicerack and use the wmflib version instead.

  • interactive: convert all imports to use the wmflib version, remove the duplicated code. The module is for now left to hold the get_management_password() function.

  • prometheus: moved entirely to wmflib.

  • _log: use the SAL (!log) IRC handler from wmflib.

  • The @retry decorator will be migrated in a separate patch to keep its dry-run awareness.

Minor improvements

• administrative: Add getters for the other Reason fields.

Bug fixes

• puppet: update get_certificate_metadata() so the pattern is more specific and prevent it to match other hosts.

• elasticsearch_cluster: fix call to @retry.

Miscellanea

• dnsdisc: improve test coverage.

• tests: fix deprecated pytest argument.

• tox: Remove --skip B322 from Bandit config not supported by newer Bandit versions.

v0.0.46 (2020-12-10)

Minor improvements

• icinga: add support for downtimed and notifications_enabled parameters (T269672).

• elasticsearch-cluster: add support for cloudelastic (T268779).

v0.0.45 (2020-11-30)

API breaking changes

• Removed config and phabricator modules migrated to wmflib and update imports.

• remote: re-enabled Cumin's output removing its suppression. The work on T212783 will make it more flexible on a per-execution basis, but for now is better to just re-enable it and make the errors surface to the users.

New features

• cookbook API: add class API

  • In addition to the simple cookbooks function API interface add support for a more integrated class-based API.

  • Spicerack will perform auto-detection of the API used by the cookbook and automatically convert the module-based API cookbooks into class-based cookbooks so that only one interface is actually supported internally.

  • The class API defines a CookbookBase class that cookbooks that want to use this API must extend creating a derived class. The derived class can have any name. Multiple cookbooks in the same module are supported.

  • The class-based API allows a more in-depth integration with Spicerack:

    • Allow to perform additional initialization and validation steps in the class constructor before the cookbook execution starts, allowing the cookbook to bail out before execution and any related !log-ging.

    • Allow to define a custom runtime description that will be included, for example, in the START/END logging messages that are also sent to IRC and !log-ed into SAL.

    • Refactor the Cookbook API documentation to be more detailed and following Sphinx standards to document the cookbooks module interfaces.

    • Refactor out from the private _cookbook module some functionalities to a _menu and _module_api modules.

• spicerack: add requests_session accessor to get a requests's Session pre-configured by wmflib with a default timeout, retry logic and User-Agent.

• decorators: Add an optional custom failure message to @retry:

  • The @retry decorator logs the messages from exceptions raised during execution, but when there are chained exceptions ("raise from", etc.) only the top-level error is logged. For example, in MediaWiki._check_siteinfo, we only log Failed to get siteinfo and throw away the message from the underlying RequestException. Instead, this traverses the exception chain (using the same logic as the built-in default handler for uncaught exceptions) and includes each exception's message in the log entry.

Minor improvements

• Convert all usage of the requests package to use the wmflib.requests.http_session instead to have a nice User-Agent, a default timeout and a retry logic on some failures across Spicerack.

• puppet: suppress deprecation warnings.

• decorators: Log chained exception messages in @retry.

Miscellanea

• doc: add missing link to the wmflib package.

• dependencies: remove temporary hacks.

• dependencies: update min version to match the versions in Debian Buster.

• tests: remove require_* decorators.

• Refactoring: renamed internal modules with a leading underscore:

  • Moved cookbook.py to _cookbook.py and log.py to _log.py as all their content is actually internal to spicerack and no client should use any of that. They were already excluded from the generated documentation for the same purpose.

v0.0.44 (2020-10-13)

Breaking changes

• dns: the dns module has been migrated to wmflib and removed from Spicerack. Its access via the spicerack.dns(() accessor is unchanged, but any direct imports from the spicerack.dns module in cookbooks must be replaced with wmflib.dns (T257905).

Miscellanea

• Spicerack now depends on the new wmflib package.

• log: adjust the return type of FilterOutCumin.filter() as required by mypy (upstream documentation incorrect).

• doc: refactor and simplify its configuration.

• pylint: allow logger as module-scope name given that is used throughout the project so that there is no need for a pylint disable comment.

v0.0.43 (2020-09-16)

Minor improvements

• elasticsearch_cluster: Store which datacenters to query for metrics in Prometheus.

v0.0.42 (2020-08-31)

Bug Fixes

• elasticsearch_cluster: fix prometheus query syntax.

v0.0.41 (2020-08-31)

Bug Fixes

• dnsdisc: change retry logic to wait up to 27 seconds with more frequent checks instead of the current 9 seconds.

v0.0.40 (2020-08-27)

New features

• elasticsearch_cluster: verify all write queues are empty querying Prometheus (T261239).

Miscellanea

• doc: improved logging documentation.

v0.0.39 (2020-08-18)

New features

• Add native mysql spicerack module.

Bug Fixes

• mysql_legacy: update Cumin queries for DB selection due to Puppet refactors.

• icinga: fix bug for recheck_all_services(), the signature of the Icinga command requires a check time too.

Miscellanea

• Remove support for Python 3.5 and 3.6.

• actions: refactored to take advantage of more recent Python versions.

• Add type hints for variables and attributes since the support for older Python versions has been dropped.

• Pin to a working version of prospector as 1.3.0 was overenthusiastic with updating its dependencies.

• actions: fix test for pytest regression in version 6.0.0.

v0.0.38 (2020-06-09)

Bug Fixes

• ganeti: update the list of available rows in the eqiad and codfw datacenters.

Miscellanea

• Add support for Python 3.8.

v0.0.37 (2020-05-18)

Bug Fixes

• icinga: fix get_status():

  • The icinga-status script that returns the status can be run also in dry-run mode as it's a read-only tool.

  • The icinga-status script exits with a non-zero exit status on non-optimal and missing hosts, accept any exit code.

v0.0.36 (2020-05-18)

Bug Fixes

• tests: add @require_caplog to some actions module tests to fix the build on Debian Stretch.

v0.0.35 (2020-05-18)

API breaking changes

• Rename mysql module to mysql_legacy:

  • The existing mysql module uses remote execution of the mysql client to interact with mysqld's. Moving this out of the way to allow room for a new mysql module which uses a native mysql client library.

New features

• interactive: add get_secret() function for requesting secrets interactively with optional ask for confirmation.

• icinga: allow to check the status of a host:

  • Add a get_status() method that allows to get the current status of a set of hosts in Icinga.

  • The returned status allow to quickly check if all the hosts are in optimal state, get a list of those that are not and the services that are failing on those hosts.

• actions: new module to track cookbook actions:

  • Add a new actions module that contains an Actions class and an ActionsDict class that is an ordered dictionary with default dictionary functionalities of Actions class instances.

  • The Actions instances allow to keep track of actions performed by acookbook with the following features:

    • Save the message of the action with different levels (success, warning, failure).

    • Log the message of the action with the associated log level.

    • Keep track of the presence of any warning or failure.

    • Have a nice string representation of the actions, suitable to be used to update a Phabricator task.

  • The ActionsDict class has too a nice string representation of its items.

  • This is a porting with some generalization of the code present in the sre.hosts.decommission cookbook.

  • Pre-create an ActionsDict instance in spicerack so that it can be accessed in the cookbooks directly as spicerack.actions.

• typing: add a typing module for custom type hints:

  • Add a new typing module to hold all custom types useful across Spicerack.

  • Define a custom type TypeHosts that can be either a NodeSet or a sequence of strings.

  • Use the new type in the icinga module.

Bug Fixes

• ipmi: fix subprocess.run() calls to raise on failure.

  • The check parameter is by default False, hence not raising an exception if the executed command exit with a non-zero exit code.

  • Forcing the check parameter to be True to ensure an exception is raised on failure.

Miscellanea

• icinga: refactor input parsing:

  • The Icinga class needs to use hostnames instead of FQDNs.

  • Move the conversion from FQDNs (or hostnames) to hostnames to a static method so that can be used across the class without repetition of code.

• tests: fix newly reported flake8 issues.

• tests: relax Prospector dependency:

  • The upstream bug that required to set an upper limit on the version of Prospector has been fixed.

  • Removing the upper bound to get newer features.

  • Fix newly reported issues.

• tests: relax Bandit dependency:

  • The upstream bug that required to set an upper limit on the version of Bandit has now a workaround using a specific syntax for the exclude files.

  • Removing the upper bound to get newer features.

  • Fix newly reported issues.

  • Remove nosec comments not needed anymore and convert some of them into skipped checks in tox.ini. This way the affected lines are still checked for other issues.

v0.0.34 (2020-05-06)

API breaking changes

• netbox: removed property device_status_choices of the Netbox class, not currently used and removed from Netbox API starting from version 2.8.0.

Bug Fixes

• netbox: adapt to new Netbox API:

  • Netbox API starting with Netbox 2.8.0 have removed the choices API endpoint. Given that it was used only for the status, removing its support completely for now given that is not directly supported by the pynetbox library yet.

Miscellanea

• doc: set min version of sphinx_rtd_theme to 0.1.9 to match Debian Stetch.

• doc: fix documentation generation for Sphinx 3.

• changelog: specify breaking change for v0.0.33.

v0.0.33 (2020-05-04)

API breaking changes

• netbox: the default instance returned when calling Spicerack.netbox() uses a read-only token. To have read-write access to Netbox the read_write parameter should be set to True.

New features

• netbox: add support for RW and RO tokens:

  • Use a RO token by default, allow to request a Netbox instance with a RW token.

  • Always use a RO token if in dry-run mode to allow to expose the Netbox API object directly to the clients.

• netbox: expose the pynetbox API object:

  • To allow to perform additional operations not yet abstracted by the Netbox class, expose the pynetbox API object directly.

  • The dry-run mode support is ensured by the RO token.

Minor improvements

• include the username in logfiles.

v0.0.32 (2020-03-11)

Minor improvements

• spicerack: allow to override Spicerack's instance parameters from the configuration file. See config.yaml.

• spicerack: allow to cache the Ipmi instance so that it can be re-used without re-asking the management password.

• spicerack: expose to cookbooks the _spicerack_config_dir parameter via a getter.

• netbox: fine tune log and exception messages.

• elasticsearch_cluster: return the cluster name in ElasticsearchCluster.__str__.

• mysql: update CORE_SECTIONS for external storage RW instances (T226704).

Bug Fixes

• elasticsearch_cluster: add https:// to relforge endpoints.

Miscellanea

• tests: remove unused mypy type ignore comments.

v0.0.31 (2020-02-26)

New features

• ganeti: add VM creation capability (T231068).

• spicerack: add support for an HTTP proxy.

  • To perform calls to external endpoints it might be necessary to use an HTTP proxy, add support for it.

  • Read the http_proxy config from the main spicerack configuration file and inject it into Spicerack that will also expose it to the cookbooks.

  • Add a getter for the http_proxy property to Spicerack.

  • Add a helper that returns a proxies dictionary to be used by the Python Requests module.

Minor improvements

• ganeti: use canonical Ganeti cluster names (T231068).

• ganeti: add logging for GntInstance actions (T231068).

v0.0.30 (2020-02-11)

API breaking changes

• netbox: rename injected property in host details (T231068).

  • When fetching host details from Netbox, Spicerack injects some properties to distinguish between virtual and physical hosts. Renaming the cluster_name property to ganeti_cluster to avoid possible confusions.

New features

• spicerack: add getter for the Netbox master host. In some cases is necessary to execute commands on the Netbox master host, add a getter to resolve its real hostname (T231068).

• ganeti: add cluster to instance() (T231068).

  • Allow to specify the Ganeti cluster name when calling instance(). If set the instance will be searched only in that cluster.

  • Pass the cluster name to the GntInstance constructor and expose it via a getter to remove the necessity to look it up separately when cluster was not passed to instance() for auto-detection.

• ganeti: add initial support for gnt-instance (T231068).

  • Add initial support for gnt-* commands to be executed on the cluster master via remote execution.

  • Add initial support for gnt-instance commands to perform Ganeti VMs decommissioning, in particular:

    • shutdown: to shutdown a Ganeti VM, with its optional timeout parameter.

    • remove: to shutdown and remove a Ganeti VM, with its optional shutdown_timeout parameter.

Minor improvements

• mediawiki: use Cumin alias instead of role query (T243935).

Miscellanea

• dnsdisc: fix typo in docstring.

v0.0.29 (2020-01-16)

Minor improvements

• mediawiki: in stop_cronjobs() adapt for the migration from hhvm to php-fpm in production (T229792).

• dnsdisc: use port 5353 to query the resolvers. The authdns part is answering to port 5353 from now on.

• dns: allow to specify a custom port for the resolver. The authdns part is answering to port 5353 from now on, allow to specify a custom port when instantiating a new Dns recursor.

• ganeti: Add esams, ulsfo and eqsin clusters and rows definitions.

Bug Fixes

• ipmi: the change introduced via I4d4ade351493a548e9e7a578bf9a7acbb45a5c0 to use subprocess.run() created a regression causing the ipmi calls to no longer capture stdout. Restored normal behaviour (T147074).

Miscellanea

• dns: remove unused type hint ignore comments.

• remote: fix docstring return type.

• doc: updated link to the requests module documentation.

• docstrings: fix pep257 reported errors.

• mypy: Get rid of no longer needed # type: ignore annotations that are now detected automatically by mypy.

v0.0.28 (2019-10-10)

Minor improvements

• netbox: Transparently support read-only operations for virtual machines (T231068).

• ganeti: Add ability to get ganeti cluster for given instance (T231068).

• ipmi: add support for channel 2.

• ipmi: use subprocess.run() instead of subprocess.check_output().

v0.0.27 (2019-08-25)

Minor improvements

• remote: Move splitting of a RemoteHosts instance to a split() method.

• netbox: Make host private and raise exception on not found.

• netbox: Add method to return host information.

v0.0.26 (2019-08-06)

New features

• Add Netbox module.

• Add the LBRemoteCluster class to manage cluster behind a load balancer.

Minor improvements

• icinga: Add a function to force a recheck of all sevices.

• confctl: Add filter_objects and update_objects.

• confctl: add change_and_revert contextmanager.

Bug Fixes

• elasticsearch_cluster: correct ports for relforge cluster.

• elasticsearch_cluster: fix mypy newly reported bug.

• tests: fix pytest caplog matching.

• tests: fix pep257 newly reported issues.

v0.0.25 (2019-05-10)

Bug Fixes

• setup.py: fix urllib3 dependency:

  • In order to build on Debian Stretch without backported packages, relax a bit the urllib3 dependency as the only goal for to specify it is to avoid conflicts with the latest version.

• doc: fix Sphinx configuration:

  • In order to avoid issues while building the Debian package on Stretch where Sphinx 1.4.9 is available, change configuration to:

    • Reduce minimum Sphinx version to 1.4.9 in setup.py.

    • Remove the warning-is-error configuration from setup.cfg that is applied to every Sphinx run, and move it directly into tox.ini as a command line -W option, that will be executed only by tox and not during the Debian package build process.

v0.0.24 (2019-05-09)

Minor improvements

• prometheus: add timeout support to query() method.

• ganeti: add timeout support.

• cookbook API: drop get_title() support:

  • No current cookbook is using the dynamic way to provide a title through get_title(args).

  • This abstraction has not proven to be useful and the fact to mangle dynamically the title of a cookbook based on the current parameter while you can then execute it with different ones doesn't seem very useful, dropping it completely from the Cookbook API.

• doc: mark Sphinx warnings as error:

  • To make the documentation building process more robust make Sphinx fail on warnings too.

  • This requires Sphinx > 1.5 and will require to use the backport version while building the package on Debian Stretch.

• doc: add checker to ensure modules are documented:

  • It's common when adding a new module to forget to add the few bits required to auto-generated its documentation.

  • Add a check to ensure that all Spicerack modules are listed in the documentation API index and that the linked files exists.

Bug Fixes

• ganeti: Fix RAPI port.

• prometheus: fix base URL template.

• doc: autodoc missing API modules.

Miscellanea

• setup.py: force urllib3 version due to pip bug.

• Add emacs ignores to gitignore.

• tests: temporarily force bandit < 1.6.0:

  • Due to a bug upstream bandit 1.6.0 doesn't honor the excluded directories, causing the failure of the bandit tox environments. Temporarily forcing its version.

v0.0.23 (2019-04-19)

New features

• Add basic Ganeti RAPI support.

• Add basic Prometheus support.

Minor improvements

• elasticsearch_cluster: add reset all indices to read/write capability (T219799).

Bug Fixes

• elasticsearch_cluster: logging during shard allocation was too verbose, some messages lowered to debug level.

Miscellanea

• flake8: enforce import order and adopt W504:

  • Add flake8-import-order to enforce the import order using the edited style that corresponds to our styleguide, see: mediawiki.org: Coding_conventions/Python.

  • Mark spicerack as local and do not specify any organization-specific packages to avoid to keep a manually curated list of packages.

  • Fix all out of order imports.

  • For line breaks around binary operators, adopt W504 (breaking before the operator) and ignore W503, following PEP8 suggestion, see: PEP0008#line_break_binary_operator.

  • Fix all line breaks around binary operators to follow W504.

v0.0.22 (2019-04-04)

Bug Fixes

• elasticsearch_cluster: use NodesGroup instead of free form JSON.

v0.0.21 (2019-04-03)

Minor improvements

• elasticsearch_cluster: Retrieve hostname and fqdn from node attributes.

• elasticsearch_cluster: make unfreezing writes more robust (T219640).

• elasticsearch_cluster: cleanup test by introducing a method to mock API calls.

• elasticsearch_cluster: rename elasticsearchclusters to elasticsearch_clusters.

Bug Fixes

• tox: fix typo in environment name.

Miscellanea

• Add Python type hints and mypy check, not for variables and properties as we're still supporting Python 3.5.

• setup.py: revert commit 3d7ab9b that forced the urllib3 version installed as it's not needed anymore.

• tests/doc: unify usage of example.com domain.

v0.0.20 (2019-03-06)

New features

• ipmi: add password reset functionality.

Minor improvements

• elasticsearch_cluster: upgrade rows one after the other.

• remote: suppress Cumin's output. As a workaround for a regression in colorama for stretch.

• Expose hostname from Reason.

• elasticsearch_cluster: use the admin Reason to get current hostname.

Bug Fixes

• debmonitor: fix missing variable for logging line.

• elasticsearch_cluster: fix typo (xarg instead of xargs).

• doc: fix reStructuredText formatting.

Miscellanea

• Drop support for Python 3.4.

• Add support for Python 3.7.

• tests: refactor tox environments.

v0.0.19 (2019-02-21)

Bug Fixes

• elasticsearch_cluster: support cluster names which have - in them.

• elasticsearch_cluster: get_next_clusters_nodes() raises ElasticsearchClusterError.

• elasticsearch_cluster: systemctl iterates explicitly on elasticsearch instances.

Miscellanea

• setup.py: add long_description_content_type.

v0.0.18 (2019-02-20)

Bug Fixes

• elasticsearch_cluster: access production clusters over HTTPS.

v0.0.17 (2019-02-20)

Minor improvements

• icinga: add remove_on_error parameter to the hosts_downtimed() context manager to decide wether to remove the downtime or not on error.

Bug Fixes

• elasticsearch_cluster: raise logging level to ERROR for elasticsearch.

• elasticsearch_cluster: retry on all urllib3 exceptions.

v0.0.16 (2019-02-18)

Bug Fixes

• elasticsearch_cluster: retry on TransportError while waiting for node to be up.

• Change !log formatting to match Stashbot expectations.

v0.0.15 (2019-02-14)

Bug Fixes

• elasticsearch_cluster: add doc type to delete query.

v0.0.14 (2019-02-13)

New features

• icinga: add context manager for downtimed hosts:

  • Add a context manager to allow to execute other commands while the hosts are downtimed, removing the downtime at the end.

• management: add management module:

  • Add a management module with a Management class to interact with the management console names.

  • For now just add a get_fqdn() method to automatically calculate the management FQDN for a given hostname.

• puppet: add check_enabled() and check_disabled() methods.

• decorators: make retry() DRY-RUN aware:

  • When running in DRY-RUN mode no real changes are done and usually the @retry decorated methods are checking for some action to be propagated or completed. Hence when in DRY-RUN mode they tend to fail and retry until the tries attempts are exhausted, adding unnecessary time to the DRY-RUN.

  • With this patch the retry() decorator is able to automagically detect if it's a DRY-RUN mode when called by any instance method that has a self._dry_run property or, in the special case of RemoteHostsAdapter derived instances, it has a self._remote_hosts._dry_run property.

• puppet: add delete() method to remove a host from PuppetDB and clean up everything on the Puppet master.

• spicerack: expose the icinga_master_host property.

• administrative: add owner getter to Reason class:

  • Add a public getter for the owner part of a reason, that retuns in a standard format the user running the code and the host where it's running.

Minor improvements

• decorators: improve tests.

• doc: fine-tune generated documentation.

Bug Fixes

• dns: remove unused dry_run argument.

• Add missing timeout to requests calls.

• dns: fix logging message.

• elasticsearch_cluster: change is_green() implementation.

• elasticsearch_cluster: fix issues found during live tests.

• spicerack: fix __version__.

• ipmi: fix typos in docstrings.

v0.0.13 (2019-01-14)

Bug Fixes

• remote: fix logging for reboot().

v0.0.12 (2019-01-10)

New features

• ipmi: add support for DRY RUN mode.

• config: add load_ini_config() function to parse INI files.

• debmonitor: use the existing configuration file:

  • Instead of requiring a new configuration file, use the existing one already setup by Puppet for the debmonitor client.

  • Inject the path of the Debmonitor config into the ctor with a default value.

• puppet: add default batch_size when running puppet:

  • Allow to specify the batch_size when running puppet on a set of hosts.

  • Add a default batch_size to avoid to overload the Puppet master hosts.

Bug Fixes

• phabricator: remove unneded pylint ignore.

• mediawiki: update maintenance host Cumin query.

• remote: add workaround for Cumin bug.

  • To avoid unnecessary waiting on the most common use case of reboot(), that is with only one host, unset the default batch_sleep as a workaround for T213296.

• puppet: fix regenerate_certificate().

  • When re-generating the certificate, Puppet will exit with status code 1 both if successful or on failure.

  • Restrict the accepted exit codes to 1.

  • Detect errors in the output and raises if any.

v0.0.11 (2019-01-08)

New features

• debmonitor: add debmonitor module.

• phabricator: add phabricator module.

Bug Fixes

• icinga: fix command_file property.

• puppet: fix subprocess call to check_output().

• dns: include NXDOMAIN in the DnsNotFound exception.

• admin_reason: fix default value for task.

v0.0.10 (2018-12-19)

API breaking changes

• cookbook: split main into argument_parser() and run().

• remote: refactor Remote.query() API.

New features

• Add administrative module.

• dns: add dns module.

• Add elasticsearch_cluster module.

• Add Icinga module.

• Add ipmi module.

• Add Puppet module.

• puppet: add additional methods to PuppetHosts.

• puppet: add PuppetMaster class.

• remote: add more host functionalities.

Minor improvements

• doc: add documentation and its generation.

• interactive: add ensure_shell_is_durable().

Bug Fixes

• administrative: fix Reason's signature.

• elasticsearch_cluster: fix tests for Python 3.5.

• icinga: fix typo in test docstring.

• interactive: check TTY in ask_confirmation().

• mediawiki: kill also HHVM on stop_cronjobs.

• Fix typo in README.rst.

• tests: fix randomly failing pylint check.

Miscellanea

• setup.py: update curator version to match our current elasticsearch version.

• setup.py: force urllib3 version.

• tests: fix lint ignore.

v0.0.9 (2018-09-12)

Minor improvements

• mediawiki: improve siteinfo checks.

• dnsdisc: improve TTL checks.

• exceptions: add SpicerackCheckError.

• tests: improve prospector tests.

Bug Fixes

• dnsdisc: catch dnspython exceptions.

• setup.py: add missing fields and fix missing comma.

v0.0.8 (2018-09-10)

Minor improvements

• mediawiki: ignore exit codes on stop_cronjobs.

• logging: minor improvements and a fix.

v0.0.7 (2018-09-06)

Bug Fixes

• dnsdisc: fix dry-run in check_if_depoolable().

v0.0.6 (2018-09-06)

Minor improvements

• log: remove relic from switchdc.

• mysql: refactor sync check to avoid GTID.

v0.0.5 (2018-09-05)

Minor improvements

• mediawiki: improve validation checks.

v0.0.4 (2018-09-04)

New features

• Add redis_cluster module.

• dnsdisc:

  • add methods for checking if a datacenter can be depooled.

  • add a pool() and depool() methods.

• mediawiki:

  • improve stop_cronjobs() method.

  • add check_cronjobs_disabled() method.

  • refactor to use confctl's set_and_verify().

  • split set_readonly() and add checks.

• mysql:

  • add get_dbs() method.

  • rename the ensure_core_masters_in_sync() method.

• confctl: add set_and_verify() method.

v0.0.3 (2018-08-30)

Miscellanea

• Change PyPI package name and add long description to setup.py.

v0.0.2 (2018-08-28)

New features

• mediawiki: add siteinfo-related methods.

v0.0.1 (2018-08-26)

• Initial version.