fatal lock wait timeout(7200) should be replaced by SRV_SEMAPHORE_WAIT_EXTENSION in row_import_tablespace_for_mysql

[yinfeng]

read code of function row_import_tablespace_for_mysql in row0mysql.c (Percona-Server-5.5.25a-rel27.1)

2704 /* Enlarge the fatal lock wait timeout during import. */
2705 mutex_enter(&kernel_mutex);
2706 srv_fatal_semaphore_wait_threshold += 7200; /* 2 hours */
2707 mutex_exit(&kernel_mutex);

actually , In MySQL5.5.25 and later version, "7200" was replaced by SRV_SEMAPHORE_WAIT_EXTENSION.

[Raghavendra D Prabhu (raghavendra-prabhu)]

SRV_SEMAPHORE_WAIT_EXTENSION is still there in PS in other places.

But the code segment quoted is no longer part of the upstream code (ie. the semaphore wait addition has been removed), however it is still part of PS.

This could either be a merge regression or a fix for upstream regression (since importing tablespaces can take longer than default 600 seconds) .

If it is an upstream regression, then srv_fatal_semaphore_wait_threshold += SRV_SEMAPHORE_WAIT_EXTENSION; would be the right fix.

[Laurynas Biveinis (laurynas-biveinis)]

Yes, the recently-introduced SRV_SEMAPHORE_WAIT_EXTENSION is a good constant to use here.

[Shahriyar Rzayev (rzayev-sehriyar)]

Percona now uses JIRA for bug reports so this bug report is migrated to: https://jira.percona.com/browse/PS-2789