The parser cache in production is using a mysql-multiwrite object cache which write to both memcached and a mysql database (pc) (see $wgParserCacheType and $wgObjectCaches['mysql-multiwrite']).
In labs, we only use the CACHE_MEMCACHED, we could use a db as well.
Probably need MariaDB puppet manifests to be adapted just like for bug 53339 (migrate databases to MariaDB).
| Status | Subtype | Assigned | Task | ||
|---|---|---|---|---|---|
| Declined | None | T55457 setup a DB backed parser cache | |||
| Resolved | None | T55458 adapt the MariaDB puppet manifests for beta |
This got filed in an attempt to make Beta-Cluster-Infrastructure closer to production. It defines an object cache that uses both memcached and mysql:
$wgObjectCaches['mysql-multiwrite'] = [
'class' => 'MultiWriteBagOStuff',
'caches' => [
0 => [
'factory' => [ 'ObjectCache', 'getInstance' ],
'args' => [ 'mcrouter' ]
],
1 => [
'class' => 'SqlBagOStuff',
'servers' => $pcServers,
'purgePeriod' => 0,
'tableName' => 'pc',
'shards' => 256,
'reportDupes' => false
],
],
'replication' => 'async',
'reportDupes' => false
];
'wgParserCacheType' => [
'default' => 'mysql-multiwrite',
'wikitech' => 'memcached-pecl',
],Where as on beta we only use memcached:
'wgParserCacheType' => [
'default' => CACHE_MEMCACHED,
],Which is probably not a big deal.