MySQL Bugs: #9822: max_allowed_packet variable changes not effective

Bug #9822	max_allowed_packet variable changes not effective
Submitted:	11 Apr 2005 16:28	Modified:	2 Jul 2005 11:15
Reporter:	Tate Baumrucker	Email Updates:
Status:	Can't repeat	Impact on me:	None
Category:	MySQL Server: Replication	Severity:	S3 (Non-critical)
Version:	4.1.11-standard-log	OS:	Linux (RH Linux 2.4.9-e.59smp AS 2.1)
Assigned to:		CPU Architecture:	Any

Description:
Replication failing with Errno: 1236 (Error reading packet from server: log event entry exceeded max_allowed_packet; Increase max_allowed_packet on master).  Have updated max_allowed_packet via my.cnf and/or via set global w/in mysql CLI on both master and slave to various sizes (including up to 1G), but continue to get slave error.  Is it possible variable is not being read properly?

How to repeat:
stop slave;
start slave;
show slave status;

just to make sure: have you changed the value on both master and slave
and have you restarted the two after changing the value in my.cnf?

Yes.  I've restarted after changing the my.cnf values ...

No feedback was provided for this bug for over a month, so it is
being suspended automatically. If you are able to provide the
information that was originally requested, please do so and change
the status of the bug back to "Open".

The problem still persists, but I cannot provide additional feedback.  Nothing has changed with the problem.  Frankly, I was hoping for freeback from you folks ...

Possibly another angle on this:

I've got a bunch of replicated servers, with max_allowed_packet set to 10M everywhere, and I'm getting errors like that as well.

I've noticed that some servers have a different idea of what "10M" is!
I checked the value on all servers like this:

  show variables like 'max_allowed_packet';

My master server gives 10484736, while some (not all!) of the slaves give 1047552...

All of them are running the same 4.1.11 binary on the same OS (debian).

Correction:

My error message is different, but the underlying problem might well be the same.

I get this on the slave:
Error reading packet from server: Got packet bigger than 'max_allowed_packet' bytes (server_errno=2020)

Setting the value explicitly to 10484736 instead of 10M on all servers solved my problem...

Hi.

Obviously it's  a workaround - to set max_allowed packet to 10484736 and do not use 10M. But we are interested in catch this bug.

make a lot of tests against 4.1.12 and 4.1.11 and was not able to reproduce this bug. Every time I got the same max_allowed_packet on master and on slave.

If you have any ideas how to reproduce it pls let us know.

The reporte have not provided what kind of query was stuck in his binlog. but i can guess:
the case resembles bug#19402 where there are queries size of max_allowed_packet in binlog.