MySQL Bugs: #19643: Replication problems

Bug #19643	Replication problems
Submitted:	9 May 2006 17:05	Modified:	11 Jun 2006 11:58
Reporter:	Keith Larson	Email Updates:
Status:	No Feedback	Impact on me:	None
Category:	MySQL Server	Severity:	S3 (Non-critical)
Version:	4.1.12-standard-log	OS:	x86_64 GNU/Linux
Assigned to:		CPU Architecture:	Any

Description:
Hello,

I am trying to determine if my problems is a bug in mysql 4.1.12 64bit. 

I have a Master DB and 3 slaves. 
All the slaves have different server_id in my.cnf. 

I just dropped all the Slaves and refreshed with new copies via IBhotbackup from the Master (A).

Started up the slaves. 

2 (B&C) of the slaves are on a local private network next to the master.  -- 64bit
1 (D) is remote sitting at my desk.   -- 32bit 

The 2 local slaves can not keep up. They slowly fall back currently both approx 45067 back. 

The little 32bit version I have can keep up or close the gap after sql errors where the 2 64bit never can.

For example: 

Server (D) had an error overnight that i repaired when I got in. 
IT was 
Seconds_Behind_Master: 60924 
Now Seconds_Behind_Master: 59251 and closing. 

Servers B & C 
This morning 
B-- Seconds_Behind_Master: 43378
C --Seconds_Behind_Master: 43963

NOW 
B-- Seconds_Behind_Master: 45098
C --Seconds_Behind_Master: 44804

As you can see they are going the wrong direction. 
Any ideas on where else to troubleshoot this?

How to repeat:
Good question. 

Start slave for now  :)

Thank you for a problem report. What Linux is used on that problematic slaves? If it is Debian, please, send the glibc version used.

Please, in any case, try to repeat with a newer version, 4.1.19.

It is Red Hat Ent. 

I have just updated the 4.1.19 on slave with no new results.
I am going to upgrade the master as well.

Please, upgrade master and inform about the results. We'll have to check on the latest version anyway.

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".