[MySQL-replication] master to master replication broke

Hi Lisa.

When you said that: " (...) 2.change master to where it was stopped", what did you do? CHANGE MASTER TO .... ?
Did you stopped the slave before executing that command?

Have you tried to delete the file master.info?


Miguel Araújo

Dear Miguel,

Lisa explicitly stated she did not want to dump/sync. You suggest:

Have you tried to delete the file master.info?
>

Kindly explain to me what this does, and what the consequences are.

Deal Johan,

Lisa said that the server was rebooted, so I guess that the master info structure was lost, and that's the reason for the server not resume replication.
She also stated the "2.change master to where it was stopped; 3. change master to current log file and position", but didn't refer if has done the "complete" CHANGE MASTER TO, command.
So I guess that the slave can't connect to the server because it lacks the IP, user, passwd or other option.

It's just a guess...

Deleting the master.info erases the master structure, and erases a possible error done in the master structure by the reboot.

Why not just login to the database and do a
show slave status;
show master status;



2010/3/17 Miguel Araújo <noradone@gmail.com>

Agreed. Lisa, please show the information given by those commands.

Bah, I seem to be on a hair trigger today. Apologies for my blut replies.

2010/3/17 Miguel Araújo <noradone@gmail.com>
Quite possible. If this isn't the case, though, she'll have lost the only
source of master log and position, and thus end up doing a full sync. Better
to verify manually first :-)


Yes, but doesn't magically set it to the right values afterwards.

No problem :-)

Yes, I have already told her to set all the right values afterwars manually.

Congratulations. By doing "3. change master to current log file and
position", you *almost* condemned yourself to the dump/sync you dread.

You're actually lucky that it didn't start replicating when you did that.
Hurry up and put it back to the original settings, before you do anything
else - if a single statement gets executed, you're screwed.

As to what's wrong, I would think "cannot connect to server" is pretty
obvious. Find out *why*. Hint: try to telnet into the IP/port configured in
your mysql slave.

Johan,

If you had read carefully, you would know the reason cannot connect to server because admin rebooted the server hosts.

I knew change master to current position is not a good solution, I just want to see if replication can be resumed. In fact, after that it is still stopped on where it was before broken. It didn't changed to new position.

Thanks,
Lisa

--- On Wed, 3/17/10, Johan De Meersman wrote:


From: Johan De Meersman <vegivamp@tuxera.be>
Subject: Re: master to master replication broke
To: database100@yahoo.com
Cc: replication@lists.mysql.com
Date: Wednesday, March 17, 2010, 10:13 AM



On Wed, Mar 17, 2010 at 4:00 PM, Database System wrote:

Hi,

I have master to master replication ran was fine. One time admin re-booted one host and followed rebooted another one. After that, one of the masters was able to automatically re-connect to master and continue replication, but another master caould not resume replication.

I did three things and all got same error message on the screen. (1. re-started slave; 2.change master to where it was stopped; 3. change master to current log file and position). The error message is:

ERROR 1201 Could not initialize master info structure; more error messages can be found in the MySQL error log.

In the error log, it only said cannot connect to sserver... Error reading relay log... read up to .... SQL thread killed ... Normal shutdown
this log is the same as in another master and that server was able to re-connect.

Any idea of how to recover the replication without dump and sync database?


Congratulations. By doing "3. change master to current log file and position", you *almost* condemned yourself to the dump/sync you dread.

You're actually lucky that it didn't start replicating when you did that. Hurry up and put it back to the original settings, before you do anything else - if a single statement gets executed, you're screwed.

As to what's wrong, I would think "cannot connect to server" is pretty obvious. Find out *why*. Hint: try to telnet into the IP/port configured in your mysql slave.

Ahh. I assumed you posted the log from after that. So, no logging whatsoever
in the error log, and show slave status just shows the 1201 ? does the
general log show anything ? Does it show master IP, user and password in the
show slave status ?

Miguel,

I just used "change master to" to reset the same log file and pos to the same as it was. If I delete master.info, what else should I do after that?

thanks,
Lisa

--- On Wed, 3/17/10, Miguel Araújo wrote:

Lisa,

Please paste the following in order for us to help you.
both my.cnf
master.info <- second master that is not replicating
mysql-log <- last 20 line is ok
show slave status;
show master status;

You should set the machine IP, user and password.
Something like: CHANGE MASTER TO MASTER_HOST='111.111.111.111', MASTER_USER='user', MASTER_PASSWORD='passwd', MASTER_LOG_FILE='mysqld-bin.000001', MASTER_LOG_POS=0;

SHOW ....   It didn't show any error.
I did those two when I reset to old position and when I reset to new position.



--- On Wed, 3/17/10, Miguel Araújo wrote:


From: Miguel Araújo <noradone@gmail.com>
Subject: Re: master to master replication broke
To: "Song Voong" <song.voong@gmail.com>
Cc: "Johan De Meersman" <vegivamp@tuxera.be>, database100@yahoo.com, replication@lists.mysql.com
Date: Wednesday, March 17, 2010, 10:35 AM



Agreed. Lisa, please show the information given by those commands.


On Mar 17, 2010, at 3:33 PM, Song Voong wrote:

Why not just login to the database and do a
show slave status;
show master status;






2010/3/17 Miguel Araújo <noradone@gmail.com>

Deal Johan,

Lisa said that the server was rebooted, so I guess that the master info structure was lost, and that's the reason for the server not resume replication.
She also stated the "2.change master to where it was stopped; 3. change master to current log file and position", but didn't refer if has done the "complete" CHANGE MASTER TO, command.
So I guess that the slave can't connect to the server because it lacks the IP, user, passwd or other option.

It's just a guess...

Deleting the master.info erases the master structure, and erases a possible error done in the master structure by the reboot.