mmm_mon will issue a flush every n seconds

[GoogleCodeExporter]

What steps will reproduce the problem?
1. Setup a master-master-slave setup.
2. Run a high insert application on the primary master.
3. Watch the inserts stall on a read lock.

What is the expected output? What do you see instead?
Not to issue a flush, and global readonly on the primary master.

What version of the product are you using? On what operating system?
Mysql 5.1.26-rc, MMM 1.0

Original issue reported on code.google.com by tribu...@gmail.com on 15 Jul 2008 at 11:39

[GoogleCodeExporter]

Hm, are you sure the flush tables with read lock was from mmmd_mon? There is 
only one
flush in the whole code base, and thats in the create_snapshot utility, which 
is only
used by mmm_get_dump, mmm_backup and mmm_clone.

As far as i know, the only thing that gets executed every n'th second on the 
active
master is "set global read_only=0".

Original comment by m...@pascalhofmann.de on 1 Nov 2008 at 9:25

[GoogleCodeExporter]

MySQL as of 5.1.15 seems to create a global read lock when changing read_only - 
even
when its OFF and should be turned OFF.

We should work around this issue in mysql_allow_write by not changing 
read_only, if
it already has the correct value.

Maybe we also should change mysql_deny_write to acquire a global read lock 
before
setting read_only=1 for MySQL prior to 5.1.15?

Original comment by m...@pascalhofmann.de on 2 Nov 2008 at 9:15

• Changed state: Accepted
• Added labels: Priority-High
• Removed labels: Priority-Medium

[GoogleCodeExporter]

Fixed in r129. Will create a new issue for Comment #2.

Original comment by m...@pascalhofmann.de on 2 Nov 2008 at 10:11

• Changed state: Fixed

[GoogleCodeExporter]

I've opened a bug report for this issue: http://bugs.mysql.com/bug.php?id=40473

Original comment by m...@pascalhofmann.de on 2 Nov 2008 at 10:54