Bug #59625 Multiple RBS break the get rseg with mininum trx_t::no code during purge

• Submitted: 20 Jan 2011 7:39
• Modified: 13 Feb 2012 3:38
• Reporter: Sunny Bains
• Status: Closed
• Category: MySQL Server: InnoDB storage engine
• Severity: S1 (Critical)
• Version: 5.5+
• OS: Any
• Assigned to: Sunny Bains
• CPU Architecture: Any

[20 Jan 2011 7:39] Sunny Bains

Description:
When writing an update UNDO log record to disk we only acquire the rseg->mutex in
trx_purge_add_update_undo_to_history(). Simply acquiring the rseg->mutex doesn't
guarantee that the undo log list will be orderd on trx_t::no (serialisation order).

How to repeat:
Code inspection while fixing bug# 59291.

[26 Jan 2011 16:19] Mikhail Izioumtchenko

could bug#55283 be a duplicate of this one?

[30 Nov 2012 9:25] Ovais Tariq

Sunny,

As I understand it this bug was discovered while working on 59291. Would this bug have to do anything with the purge lag seen in 59291.

[2 Dec 2012 8:22] Sunny Bains

This bug has got nothing to do with purge lag, at least I can't see how.