Comments on: Making replication a bit more reliable

By: Shlomi Noach

Shlomi Noach — Mon, 16 Nov 2009 10:53:28 +0000

@Aric,

Skipping errors by specifying “slave-skip-errors = all” in my.cnf is quite dangerous, since you will not even be aware of replication issues; with Ernesto’s solution, you are at least forced into manual intervention, which may supply you with enlightenment about the problem’s origins.

Instead, your replicated data will keep on moving away from master’s data.

By: Baron Schwartz

Baron Schwartz — Sun, 15 Nov 2009 15:21:30 +0000

These “solutions” make replication LESS reliable and MORE likely to have further problems! FIX the problem, don’t hide it.

By: Aric

Aric — Sun, 15 Nov 2009 07:17:12 +0000

Also, in my.cnf you can skip replication errors.

slave-skip-errors = all

By: Alex P.

Alex P. — Wed, 18 Mar 2009 00:41:39 +0000

thank you. i will try it and find if can solve the problem.

By: Ernesto Vargas

Ernesto Vargas — Tue, 10 Mar 2009 06:24:29 +0000

Alex P,

You could also do:

STOP SLAVE; SET GLOBAL SQL_SLAVE_SKIP_COUNTER=1; START SLAVE;

So the replication will jump that Dup Key error and continue with the next statement of the replication.

By: Alex P.

Alex P. — Fri, 06 Mar 2009 18:46:50 +0000

yes i know the “duplicate key error” may be related to different issues (sometimes its isp problem, manual inserts, our slaves uses dyndns and the update of the service sometime make it crash, etc ), the way i resolve the crash is making a backup of the master, send this backup to the slave (via sftp) then i execute this backup, delete the ‘relay’ files and restart the mysql service… after that slave is runnning ok again. I know this “solution” is not very good (what do you think? ) so im trying to find another one, that’s why i think this patch/feature will help me.

By: Vadim

Vadim — Fri, 06 Mar 2009 18:05:36 +0000

Alex P, Gil

Actually this patch can fix that issue.

The problem is when your slave crash or some another problem – relay-log.info contains old information, and when you do slave start – slave tries to repeat transactions which were already executed (and you getting Duplicate key error).

Our patch tries to fix that and change relay-log.info to the real last transaction.

However Dublicate key error may be related to different issues also, so I can’t say for sure our patch is solution.

By: Gil

Gil — Fri, 06 Mar 2009 17:03:42 +0000

Alex, I have run into similar duplicate key errors in the past. It turns out the slave I was using to take backups had slightly different data than the master. You may want to do a consistency check using Maatkit just to be sure.

By: Alex P.

Alex P. — Fri, 06 Mar 2009 01:06:29 +0000

hi. Sometimes i got (in slave) the “Duplicate key entry” error (i see this in the ’show slave status’ Last_error) always with an insert query. Dont know exactly why but just happened in 1 of 3 slave, and the others was fine and running normally. With Myisam just need “load data from master” but with InnoDB we have to download and run a backup of the Master DB and it takes some minutes….. i have this problem around 1 time every 3 months but im looking for a better solution..
thanks again.

By: Vadim

Vadim — Thu, 05 Mar 2009 20:30:27 +0000

Alex,

It may help, or may not. To be sure I need to know what exact problem you have with restoring replication.