InnoDB in crash recovery initialises changed page tracking too late

[Laurynas Biveinis]

In 5.7, if crash recovery needs to be run, and changed page tracking is enabled:

...
2016-08-12T09:10:11.293017Z 0 [Note] InnoDB: Completed initialization of buffer pool
2016-08-12T09:10:11.310452Z 0 [Note] InnoDB: Recovering partial pages from the parallel doublewrite buffer at /Users/laurynas/percona/obj-bug1612551-5.7-debug-asan/mysql-test/var/mysqld.1/data/xb_doublewrite
2016-08-12T09:10:11.316683Z 0 [Note] InnoDB: Highest supported file format is Barracuda.
2016-08-12T09:10:11.318051Z 0 [Note] InnoDB: Log scan progressed past the checkpoint lsn 29278081
2016-08-12T09:10:11.318224Z 0 [Note] InnoDB: Doing recovery: scanned up to log sequence number 29336044
2016-08-12T09:10:11.337992Z 0 [Note] InnoDB: Doing recovery: scanned up to log sequence number 29336044
2016-08-12T09:10:11.353669Z 0 [Note] InnoDB: Database was not shutdown normally!
2016-08-12T09:10:11.353709Z 0 [Note] InnoDB: Starting crash recovery.
2016-08-12T09:10:11.400959Z 0 [Note] InnoDB: Created parallel doublewrite buffer at /Users/laurynas/percona/obj-bug1612551-5.7-debug-asan/mysql-test/var/mysqld.1/data/xb_doublewrite, size 3932160 bytes
2016-08-12T09:10:11.956773Z 0 [ERROR] InnoDB: The age of the oldest untracked record exceeds the log group capacity!
2016-08-12T09:10:11.956835Z 0 [ERROR] InnoDB: Stopping the log tracking thread at LSN 0
2016-08-12T09:10:11.985494Z 0 [Note] InnoDB: Starting an apply batch of log records to the database...
...

The error message "Stopping the log tracking thread at LSN 0" indicates a log write attempt before changed page tracking was initialised: the thread is not running yet, and tracked LSN 0 is uninitialised value.

[Laurynas Biveinis (laurynas-biveinis)]

     https://github.com/percona/percona-server/pull/821

[Shahriyar Rzayev (rzayev-sehriyar)]

Percona now uses JIRA for bug reports so this bug report is migrated to: https://jira.percona.com/browse/PS-1731