Hi all-
I'm configuring a Xen domU RA in Pacemaker with allow-migrate=true in a 2
node cluster. Migrations work just fine when two nodes are up and running.
Ideally, I'd like the VM to fail-over (non-live, obviously) to node2 if
node1 goes down unexpecedly. In my tests, this works OK but it takes quite
a while before Xen fires up the VM on the remaining node. I've found it
takes about a 1m to 1.5m before the RA boots up the VM on the remaining
node. It seems it's waiting on the live migration to timeout even though
its not possible with 1 node? If I disable live migration, fail over works
quickly. I've found references to the migrate-to and migrate-from timeouts
but cannot seem to find any documentation on these. Is it possible to
manipulate these in a way that will ensure VMs are booted up quickly in the
event of node failure while insuring live migrations have enough time to
complete when executed manually?
Ive included an example of a resource I am using. I found I needed to
increase the start timeout to 90s or resources would timeout when starting
after failover.
Thanks in advance
C-SMITH
<primitive class="ocf" id="VM4" provider="heartbeat" type="Xen">
<instance_attributes id="VM4-i_a">
<nvpair id="vm4-attr-01" name="xmfile" value="/xen/vm/vm4.cfg"/>
<nvpair id="vm4-attr-02" name="target-role" value="started"/>
<nvpair id="vm4-attrrr" name="allow-migrate" value="true"/>
</instance_attributes>
<operations id="vm4-ops">
<op id="vm4-monitor-10s" interval="10s"
name="monitor" timeout="60s"/>
<op id="vm4-start-30s" name="start" interval="0s"
timeout="90s"/>
<op id="vm4-stop-600s" name="stop" interval="0s"
timeout="300s"/>
<op id="vm4-mig" name="migrate-to" interval="0s"
timeout="25s"/>
</operations>
</primitive>
_______________________________________________
Linux-HA mailing list
[email protected]
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems
