Bug #53 In some cases, the slave deletes relay logs while he still may need them
Submitted: 31 Jan 2003 22:56 Modified: 24 Apr 2003 14:02
Reporter: Guilhem Bichot Email Updates:
Status: Closed Impact on me:
None 
Category:MySQL Server: Replication Severity:S3 (Non-critical)
Version:4.0, 4.1.0 OS:Any (all)
Assigned to: Guilhem Bichot CPU Architecture:Any

[31 Jan 2003 22:56] Guilhem Bichot 
Description:
I just observed this.

How to repeat:
Run replication with many transactions, and a small max_binlog_size on the slave, so that transactions get cut between several relay logs.
While it's running, do STOP SLAVE, SHOW SLAVE STATUS, and see how the printed relay log does not exist on the disk.

Suggested fix:
I'm working on it.
[31 Jan 2003 23:59] MySQL Developer 
Not exact, the bug is a bit different. When I do STOP SLAVE, and SHOW SLAVE STATUS, the printed relay log exists, but it's the relay log than contains the end of the transaction. The previous one, which contains the beginning, has been deleted. But at next START SLAVE time, to replay the transaction, the slave would need the deleted log.
[3 Mar 2003 6:48] Guilhem Bichot 
need a bit of time to finish testing the patch (always delayed because of other bugs, support questions, San Jose blah blah)
[11 Mar 2003 6:04] Guilhem Bichot 
I have a patch but it changes many lines so it may go into 4.1, not 4.0.
[8 Apr 2003 9:51] Guilhem Bichot 
will be fixed in 4.1.1
[22 Apr 2003 4:35] Guilhem Bichot 
waiting for Monty's approval before applying a patch to 4.1
(4.0 won't be fixed).
[24 Apr 2003 14:02] Guilhem Bichot 
Thank you for your bug report. This issue has been fixed in the latest
development tree for that product. You can find more information about
accessing our development trees at 
    http://www.mysql.com/doc/en/Installing_source_tree.html

Fixed in 4.1.1 (cset 1.1561) and documented.