I'm attempting to switch to row-based replication but cannot get it
(either default MIXED mode or forced ROW) to work. I'm having two
show-stopper problems when RBR is active that I don't have at all when
it is not:
1. Replication is heinously slow because it seems to be CPU bound.
It's constantly in the "invalidating query cache entries (table)"
thread state even though I have query_cache_type = 0 (i.e., OFF). I
see mysql cranking away with 99% CPU usage, whereas when I use
statement-based replication it is decidely IO bound, with barely any
CPU usage that isn't IO wait. I had hoped for the opposite, and this
seems like buggy behavior...
2. I can't stop and start the replication at all without breaking the
replication. If I do "stop slave" and then "start slave" again it
immediately has an error like this, seemingly for random tables that
have been written to (not one in particular):
Could not execute Write_rows event on table
XXX; Duplicate entry
'YYY' for key 'ZZZ',
Error_code: 1062; handler error HA_ERR_FOUND_DUPP_KEY; the event's
master log binlog.000003, end_log_pos 704849
I have scoured the bug reports and found nothing to explain either of
these. Does anyone have any ideas? I'm running the Linux (AMD64 /
Intel EM64T) build of 5.1.25 on a xeon box...nothing that unusual
about my environment or database. Only thing I could think of is that
my app does some large insert delayed ... values ... statements. If
nobody has ideas I suppose I will file bug reports, but these are such
general failures I don't know what to say in them other than "it don't
work".
thanks,
eric
--
http://ir.iit.edu/~ej--
MySQL Replication Mailing List
For list archives:
http://lists.mysql.com/replicationTo unsubscribe:
http://lists.mysql.com/replication?unsub=lists@...
