Code Coverage |
||||||||||
Lines |
Functions and Methods |
Classes and Traits |
||||||||
| Total | |
20.55% |
15 / 73 |
|
0.00% |
0 / 7 |
CRAP | |
0.00% |
0 / 1 |
| ElasticaConnection | |
20.83% |
15 / 72 |
|
0.00% |
0 / 7 |
309.79 | |
0.00% |
0 / 1 |
| getServerList | n/a |
0 / 0 |
n/a |
0 / 0 |
0 | |||||
| getMaxConnectionAttempts | |
0.00% |
0 / 1 |
|
0.00% |
0 / 1 |
2 | |||
| setTimeout | |
0.00% |
0 / 4 |
|
0.00% |
0 / 1 |
6 | |||
| setConnectTimeout | |
0.00% |
0 / 4 |
|
0.00% |
0 / 1 |
6 | |||
| getClient | |
28.30% |
15 / 53 |
|
0.00% |
0 / 1 |
75.29 | |||
| getIndex | |
0.00% |
0 / 1 |
|
0.00% |
0 / 1 |
2 | |||
| getIndexName | |
0.00% |
0 / 7 |
|
0.00% |
0 / 1 |
20 | |||
| destroyClient | |
0.00% |
0 / 2 |
|
0.00% |
0 / 1 |
2 | |||
| 1 | <?php |
| 2 | |
| 3 | namespace MediaWiki\Extension\Elastica; |
| 4 | |
| 5 | use Elastica\Client; |
| 6 | use Elastica\Index; |
| 7 | use Mediawiki\Http\Telemetry; |
| 8 | use MediaWiki\Logger\LoggerFactory; |
| 9 | |
| 10 | /** |
| 11 | * Forms and caches connection to Elasticsearch as well as client objects |
| 12 | * that contain connection information like \Elastica\Index. Propagates |
| 13 | * distributed tracing headers. |
| 14 | * |
| 15 | * This program is free software; you can redistribute it and/or modify |
| 16 | * it under the terms of the GNU General Public License as published by |
| 17 | * the Free Software Foundation; either version 2 of the License, or |
| 18 | * (at your option) any later version. |
| 19 | * |
| 20 | * This program is distributed in the hope that it will be useful, |
| 21 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 22 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
| 23 | * GNU General Public License for more details. |
| 24 | * |
| 25 | * You should have received a copy of the GNU General Public License along |
| 26 | * with this program; if not, write to the Free Software Foundation, Inc., |
| 27 | * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA. |
| 28 | * http://www.gnu.org/copyleft/gpl.html |
| 29 | */ |
| 30 | abstract class ElasticaConnection { |
| 31 | /** |
| 32 | * @var ?Client |
| 33 | */ |
| 34 | protected $client; |
| 35 | |
| 36 | /** |
| 37 | * @return string[] server ips or hostnames |
| 38 | */ |
| 39 | abstract public function getServerList(); |
| 40 | |
| 41 | /** |
| 42 | * How many times can we attempt to connect per host? |
| 43 | * |
| 44 | * @return int |
| 45 | */ |
| 46 | public function getMaxConnectionAttempts() { |
| 47 | return 1; |
| 48 | } |
| 49 | |
| 50 | /** |
| 51 | * Set the client side timeout to be used for the rest of this process. |
| 52 | * @param int $timeout timeout in seconds |
| 53 | */ |
| 54 | public function setTimeout( $timeout ) { |
| 55 | $client = $this->getClient(); |
| 56 | // Set the timeout for new connections |
| 57 | $client->setConfigValue( 'timeout', $timeout ); |
| 58 | foreach ( $client->getConnections() as $connection ) { |
| 59 | $connection->setTimeout( $timeout ); |
| 60 | } |
| 61 | } |
| 62 | |
| 63 | /** |
| 64 | * Set the client side connect timeout to be used for the rest of this process. |
| 65 | * @param int $timeout timeout in seconds |
| 66 | */ |
| 67 | public function setConnectTimeout( $timeout ) { |
| 68 | $client = $this->getClient(); |
| 69 | // Set the timeout for new connections |
| 70 | $client->setConfigValue( 'connectTimeout', $timeout ); |
| 71 | foreach ( $client->getConnections() as $connection ) { |
| 72 | $connection->setConnectTimeout( $timeout ); |
| 73 | } |
| 74 | } |
| 75 | |
| 76 | /** |
| 77 | * Fetch a connection. |
| 78 | * @return Client |
| 79 | */ |
| 80 | public function getClient() { |
| 81 | if ( $this->client === null ) { |
| 82 | // Setup the Elastica servers |
| 83 | $servers = []; |
| 84 | $serverList = $this->getServerList(); |
| 85 | if ( !is_array( $serverList ) ) { |
| 86 | $serverList = [ $serverList ]; |
| 87 | } |
| 88 | foreach ( $serverList as $server ) { |
| 89 | if ( !is_array( $server ) ) { |
| 90 | $server = [ 'host' => $server ]; |
| 91 | } |
| 92 | $server['headers'] = ( $server['headers'] ?? [] ) |
| 93 | + Telemetry::getInstance()->getRequestHeaders(); |
| 94 | $servers[] = $server; |
| 95 | } |
| 96 | |
| 97 | $this->client = new Client( [ 'servers' => $servers ], |
| 98 | /** |
| 99 | * Callback for \Elastica\Client on request failures. |
| 100 | * @param \Elastica\Connection $connection The current connection to elasticasearch |
| 101 | * @param \Exception $e Exception to be thrown if we don't do anything |
| 102 | * @param \Elastica\Client $client |
| 103 | */ |
| 104 | function ( $connection, $e, $client ) { |
| 105 | // We only want to try to reconnect on http connection errors |
| 106 | // Beyond that we want to give up fast. Configuring a single connection |
| 107 | // through LVS accomplishes this. |
| 108 | if ( !( $e instanceof \Elastica\Exception\Connection\HttpException ) ) { |
| 109 | LoggerFactory::getInstance( 'Elastica' ) |
| 110 | ->error( 'Unknown connection exception communicating with Elasticsearch: {class_name}', |
| 111 | [ 'class_name' => get_class( $e ) ] ); |
| 112 | return; |
| 113 | } |
| 114 | if ( $e->getError() === CURLE_OPERATION_TIMEOUTED ) { |
| 115 | // Timeouts shouldn't disable the connection and should always be thrown |
| 116 | // back to the caller so they can catch it and handle it. They should |
| 117 | // never be retried blindly. |
| 118 | $connection->setEnabled( true ); |
| 119 | throw $e; |
| 120 | } |
| 121 | if ( $e->getError() === CURLE_PARTIAL_FILE ) { |
| 122 | // This means the connection dropped before the full response was read, |
| 123 | // likely some sort of network problem or elasticsearch shut down |
| 124 | // mid-response. If the network failed or elasticsearch is gone the |
| 125 | // retry should fail, but we delegate deciding on retries to the caller. |
| 126 | LoggerFactory::getInstance( 'Elastica' ) |
| 127 | ->error( 'Error communicating with elasticsearch, connection closed' . |
| 128 | 'before full response was read.', [ 'exception' => $e ] ); |
| 129 | $connection->setEnabled( true ); |
| 130 | throw $e; |
| 131 | } |
| 132 | if ( $e->getError() !== CURLE_COULDNT_CONNECT ) { |
| 133 | LoggerFactory::getInstance( 'Elastica' ) |
| 134 | ->error( 'Unexpected connection error communicating with Elasticsearch. ' . |
| 135 | 'Curl code: {curl_code}', [ 'curl_code' => $e->getError() ] ); |
| 136 | // If there are different connections we could try leave this connection disabled |
| 137 | // and let Elastica retry on a different connection. |
| 138 | if ( $client->hasConnection() ) { |
| 139 | return; |
| 140 | } |
| 141 | // Otherwise this was the last available connection. Re-enable it but throw |
| 142 | // so that retries are delegated to the application. This prevents the |
| 143 | // situation where the calling code knows it can retry but no connections remain. |
| 144 | $connection->setEnabled( true ); |
| 145 | throw $e; |
| 146 | } |
| 147 | // Keep track of the number of times we've hit a host |
| 148 | static $connectionAttempts = []; |
| 149 | $host = $connection->getParam( 'host' ); |
| 150 | $connectionAttempts[ $host ] = isset( $connectionAttempts[ $host ] ) |
| 151 | ? $connectionAttempts[ $host ] + 1 : 1; |
| 152 | |
| 153 | // Check if we've hit the host the max # of times. If not, try again |
| 154 | if ( $connectionAttempts[ $host ] < $this->getMaxConnectionAttempts() ) { |
| 155 | LoggerFactory::getInstance( 'Elastica' ) |
| 156 | ->info( "Retrying connection to {elastic_host} after {attempts} attempts", |
| 157 | [ |
| 158 | 'elastic_host' => $host, |
| 159 | 'attempts' => $connectionAttempts[ $host ], |
| 160 | ] ); |
| 161 | $connection->setEnabled( true ); |
| 162 | } elseif ( !$client->hasConnection() ) { |
| 163 | // Don't disable the last connection, but don't let it auto-retry either. |
| 164 | $connection->setEnabled( true ); |
| 165 | throw $e; |
| 166 | } |
| 167 | } |
| 168 | ); |
| 169 | } |
| 170 | |
| 171 | return $this->client; |
| 172 | } |
| 173 | |
| 174 | /** |
| 175 | * Fetch the Elastica Index. |
| 176 | * @param string $name get the index(es) with this basename |
| 177 | * @param string|bool $type type of index (named type or false to get all) |
| 178 | * @param mixed $identifier if specified get the named identifier of the index |
| 179 | * @param bool $altIndex get an alternative index |
| 180 | * @param int $altIndexId get an alternative index with this ID |
| 181 | * @return Index |
| 182 | */ |
| 183 | public function getIndex( $name, $type = false, $identifier = false, bool $altIndex = false, int $altIndexId = 0 ) { |
| 184 | return $this->getClient()->getIndex( $this->getIndexName( $name, $type, $identifier, $altIndex, $altIndexId ) ); |
| 185 | } |
| 186 | |
| 187 | /** |
| 188 | * Get the name of the index. |
| 189 | * @param string $name get the index(es) with this basename |
| 190 | * @param string|bool $type type of index (named type or false to get all) |
| 191 | * @param mixed $identifier if specified get the named identifier of the index |
| 192 | * @param bool $altIndex get an alternative index |
| 193 | * @param int $altIndexId get an alternative index with this ID |
| 194 | * @return string name of index for $type and $identifier |
| 195 | */ |
| 196 | public function getIndexName( |
| 197 | $name, |
| 198 | $type = false, |
| 199 | $identifier = false, |
| 200 | bool $altIndex = false, |
| 201 | int $altIndexId = 0 |
| 202 | ) { |
| 203 | if ( $type ) { |
| 204 | $name .= '_' . $type; |
| 205 | } |
| 206 | if ( $altIndex ) { |
| 207 | $name .= "_alt_" . $altIndexId; |
| 208 | } |
| 209 | if ( $identifier ) { |
| 210 | $name .= '_' . $identifier; |
| 211 | } |
| 212 | return $name; |
| 213 | } |
| 214 | |
| 215 | public function destroyClient() { |
| 216 | $this->client = null; |
| 217 | ElasticaHttpTransportCloser::destroySingleton(); |
| 218 | } |
| 219 | } |
| 220 | |
| 221 | class_alias( ElasticaConnection::class, 'ElasticaConnection' ); |