[Pacemaker] iscsi migration to slow (disk errors) tune pacemaker or open-iscsi?

Vladislav Bogdanov bubble at hoster-ok.com
Thu Feb 24 13:51:27 UTC 2011


Hi,

sorry for top-posting.

I'd tune node.session.timeo.replacement_timeout,
node.conn[0].timeo.noop_out_interval and
node.conn[0].timeo.noop_out_timeout.

I use (but I use two portals and multipath as well, this imposes a bit
different handling) 360, 0, 0 for them.

I use iet at target side, so I also need (at least needed when I first
tried with iet 1.4.18) to block incoming/outgoing iscsi traffic with
iptables on target while migrating target/lun/ips to another node. tgt
seems (at least what I read in various places around intenet) to be free
from this issue, so you probably do not need this.

So, tuning that three parameters should be enough. First can be set to
some very high value (3600 or even much bigger), so IO requests will be
blocked for that time, IO on VM will freeze, but you will not get
read-only root partition there.

Hope this helps,

Vladislav


24.02.2011 14:05, Jelle de Jong wrote:
> Hello everybody,
> 
> I started a nice new tread for this,
> 
> I seem to got a working drbd/iscsi pacemaker setup that behaves as
> expected when nodes go done/up or resources are moved. (only got some
> issues with the high verbosity in syslog, but that is unrelated to this
> thread)
> 
> When resources are migrated on the crm nods, the open-iscsi client side
> will get disk errors and the running kvm quest will get in trouble, this
> only seems to happen when there is a lot of disk activity on the kvm
> quest, (higher loads on all systems)
> 
> # summary syslog of problem and pacemaker and iscsi configurations.
> http://debian.pastebin.com/FM3P0Qe8
> 
> Would somebody give some advice on either optimising my pacemaker
> configuration or tune my open-iscsi (timeouts?) some examples or good
> documentation URL's would be appreciated.
> 
> With kind regards,
> 
> Jelle de Jong
> 
> _______________________________________________
> Pacemaker mailing list: Pacemaker at oss.clusterlabs.org
> http://oss.clusterlabs.org/mailman/listinfo/pacemaker
> 
> Project Home: http://www.clusterlabs.org
> Getting started: http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf
> Bugs: http://developerbugs.linux-foundation.org/enter_bug.cgi?product=Pacemaker





More information about the Pacemaker mailing list