--innodb_purge_threads=1 may result in incomplete purge with innodb_fast_shutdown=0

[Laurynas Biveinis]

A 5.5-only issue, the main loop of the separate 5.5 purge thread has the following break condition:

  /* Check for shutdown and whether we should do purge at all. */
  if (srv_force_recovery >= SRV_FORCE_NO_BACKGROUND
      || srv_shutdown_state != 0
      || srv_fast_shutdown) {

   break;
  }

Which means immediate exit on srv_shutdown_state != 0, even if the thread should continue to purge if srv_fast_shutdown == 0.

This is likely the cause of bug 756387. This is an upstream issue but the upstream defaults to --innodb-purge-threads=0 whereas the PS default is 1.

[Laurynas Biveinis (laurynas-biveinis)]

     https://github.com/percona/percona-server/pull/769
     https://github.com/percona/percona-server/pull/770
     https://github.com/percona/percona-server/pull/771

[Shahriyar Rzayev (rzayev-sehriyar)]

Percona now uses JIRA for bug reports so this bug report is migrated to: https://jira.percona.com/browse/PS-3506