<div class="gmail_quote">On 29 May 2012 17:19, Stallmann, Andreas <span dir="ltr"><<a href="mailto:AStallmann@conet.de" target="_blank">AStallmann@conet.de</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
Hi!<br>
<br>
> MySQL has changed CHANGE MASTER TO syntax in 5.1 (IIRC), and it won't accept<br>
> an empty host argument anymore. I had to manually patch the RA to use '-' as the host argument if it's empty.<br>
<br>
Did you send your patch to Yves? Yves, did you include this patch in the newest resource agent?<br>
<br>
Cheers,<br>
<br>
Andreas<br>
PS: Yves, would you consider your mysql resource agent to be "production stable"? We're currently on our way to switch from mysql on drbd to mysql replication and I'm asking myself if this is a wise idea, yet. No offence meant; I'm just curious what your opinion is.<br>
</blockquote><div><br></div><div>lately it seems to work fine on my test VMs (mysql master-slave, haproxy balancing reads, as a backend for magento; siege misses a few requests while the failover occurs, the slave becomes master, the former master becomes a slave after it comes back online and receives a reader IP only after it catches up, as expected)</div>
<div><br></div><div>however, I did see some things that can be categorised as "bugfoot"[1]: it set master pref to 1 for both machines once, and another time it set the slave log file and position wrong, irrespective of the cluster property.</div>
<div><br></div><div>I just sent a pull request with some improvements, btw: <a href="https://github.com/ClusterLabs/resource-agents/pull/98">https://github.com/ClusterLabs/resource-agents/pull/98</a></div><div><br></div><div>
[1] bugfoot: a bug seen by a single person once, and they swear it was real</div><div><br></div><div>hth</div></div>