Bug #71055 | Using IF EXISTS(SELECT * ...) acquires a lock when using read uncommitted | ||
---|---|---|---|
Submitted: | 3 Dec 2013 5:18 | Modified: | 9 Dec 2013 3:00 |
Reporter: | Yoseph Phillips | Email Updates: | |
Status: | Verified | Impact on me: | |
Category: | MySQL Server: Locking | Severity: | S2 (Serious) |
Version: | 5.5.12, 5.5.35, 5.6.16 | OS: | Windows |
Assigned to: | CPU Architecture: | Any | |
Tags: | lock, read uncommitted, regression |
[3 Dec 2013 5:18]
Yoseph Phillips
[3 Dec 2013 17:26]
Sveta Smirnova
Thank you for the report. Verified as described. Problem is not repeatable with 5.1
[3 Dec 2013 17:35]
Sveta Smirnova
Workaround with --innodb_locks_unsafe_for_binlog=1 does not work for me.
[3 Dec 2013 17:37]
Sveta Smirnova
test case for MTR
Attachment: bug71055.test (application/octet-stream, text), 455 bytes.
[3 Dec 2013 17:37]
Sveta Smirnova
option file for the test
Attachment: bug71055-master.opt (application/octet-stream, text), 108 bytes.
[9 Dec 2013 3:00]
Yoseph Phillips
Just confirming what Sveta has written: the innodb_locks_unsafe_for_binlog workaround it not working for us either. (I had found that from a different post for dealing with read uncommitted bugs). Currently I have switched the tables causing the biggest performance problems from InnoDB to MyISAM, however we cannot do this for all of our tables, so this is still a big problem for us for many of our tables, so we still do not have a viable workaround.