Hello,

i'm having a really strange problem with mysql replication. I want to
upgrade two servers from MySQL 4.1.21 to 5.0.45 (which is the
current default in SuSE 10.3). Tables are a mix of InnoDB and MyISAM
tables.

I've upgraded the master, restored an SQL dump, that was made with
4.1.21 and set up replication with "change master to ..." and started the
slave. So far everything went fine, but now the trouble began.

As soon as the slave reaches the end of a relay-log file and switch to the
next one, it seems to execute a rollback command. That lead to the
situation, that the MyISAM tables are OK, but all InnoDB tables are rolled
back to the state that it had at the beginning of the relay-log.
Autocommit is set to "1" so normally it should commit after every single
update (as long as there is no explicit begin and commit/rollback).

That problem start to drive me crazy. Does anyone have a clue why the
replication act this way and how to solve that problem?

Thanks
Stefan

Hello again...

> I've upgraded the master, restored an SQL dump, that was made with
> 4.1.21 and set up replication with "change master to ..." and started the
> slave. So far everything went fine, but now the trouble began.

Well there was a mistake. Of course i've upgraded the slave, not the
master. Sorry for the confusion.