R: how use sql_slave_skip_counter to restore slave replication

[AESYS S.p.A. [Enzo Arlati]]

I already tried the option --slave-skip-errors, but i didn't work as
expected.
When this option is enabled, the slave didn't report the error message, but
the slave is stopped due the error.

The target I need to reech is to get two server A e B, each of them have its
own db server and a set of application on top of it.
Only one server should be active ( suppose A ), while on the other (B), an
application detect that A is already running and configure at run-time B as
slave.
When A crash or for some other reason stop working, an application on B ,
which loose the communication with A, configure B as master.
When A restart, it check B , and verify that B is the master, so it
configure itself as slave.
All these should work automatically, that is without manual operations.

For this application the critical point is that one of this server must
always be running.
If I loose a little amount of data (mainly log data), this should be
considered a little problem.


What I have to resolve is the lots of duplicate data I receive.
Also this is not clear, because I should undertand that if both the server
for some reason works at the same time for a while but I can't undertand
while the same error for the same keys happened so many times.
It seems that somewhere remains this commands which are aplways resubmitted,
also if I clear both master and slave.


-----Messaggio originale-----
Da: sheeri kritzer [mailto:awfief@stripped]
Inviato: lunedì 9 gennaio 2006 19.46
A: mysql@stripped
Oggetto: Re: how use sql_slave_skip_counter to restore slave replication


(still catching up, sorry about the old issues)

Don't know what version is being used here, but one thing you might
want to use, if you're just going to skip the errors anyway (hopefully
you understand WHY you're getting them), is to start replication with
the

--slave-skip-errors

option -- you can skip only the errors you want or all errors.

See:
http://dev.mysql.com/doc/refman/5.1/en/replication-options.html

I'd recommend getting replication up to speed with the
slave-skip-errors option, then stopping the slave, taking out that
option, so that new errors will show up.

.....................

I'm also going to guess that the original poster did not lock all
tables for the duration of the backup, getting a *snapshot* backup of
the database, and that's why the replication errors happened.

-Sheeri