[Labs-l] Where can I see the Tool Labs DB replag?

[Asher Feldman]

On Fri, Jun 14, 2013 at 1:46 AM, Tim Landscheidt <tim at tim-landscheidt.de>wrote:

> Asher Feldman <afeldman at wikimedia.org> wrote:
>
> > Replication was halted on a lock timeout, likely caused by a query
> writing
> > to a userdb while reading from a dewiki table.  I had intended to set
> > innodb_locks_unsafe_for_binlog on all labsdb servers, which disables gap
> > and next key locking, while also only guaranteeing read_committed
> > isolation.  This is completely safe in labs since nothing but the slave
> sql
> > thread writes to the replicated tables, and it should prevent this issue.
> > I ended up setting it for everything except s4 and s5; probably related
> to
> > the mad scrambling for the Amsterdam hackathon :)  S4 and s5 now need a
> > mariadb restart for the setting to go live, which I'll work with Marc to
> > schedule.
>
> > I've also set slave_transaction_retries to a very large value on all of
> the
> > labsdb's, a change which does not require a restart.  That should also
> > prevent a recurrence in the interim before the the server restarts occur.
> > And at least the s5 replica caught up very quickly.
>
> > [...]
>
> Is this limited to the replicated tables or does it affect
> the user tables as well?  Does this mean REPEATABLE READ is
> not available at all on the labsdb servers?
>

Essentially.  While tx_isolation is set to REPEATABLE READ, only READ
COMMITTED is actually guaranteed.  This applies to all labsdb systems.


> Tim
>
>
> _______________________________________________
> Labs-l mailing list
> Labs-l at lists.wikimedia.org
> https://lists.wikimedia.org/mailman/listinfo/labs-l
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.wikimedia.org/pipermail/labs-l/attachments/20130618/2b5f2b2d/attachment.html>