[Pacemaker] monitoring attributes

Giuliano Zorzi g.zorzi at telerix.it
Mon Mar 3 10:12:44 UTC 2014


Il 03/03/2014 10:45, Michael Schwartzkopff ha scritto:
> Am Montag, 3. März 2014, 10:26:30 schrieb Giuliano Zorzi:
>> Hi all,
>> I'm stuck on something I thought to be very simple, but googling hasn't
>> helped me :(
>>
>> I have a 2 node cluster with an IPaddress2 resource and a
>> pacemaker:pingd clone. The resource goes away from the node with pingd
>> value = 0 just fine.
>>
>> Now I need to run a script on the "failed" node when it comes back but I
>> can't seem to find a way.
> No. You make your resources depend on the result of the ping monitoring by a
> location constraint and the score_attribute "pingd". So the ping resource
> checks the availabilily of the hosts in the list and updates the result in the
> pingd attribute of avery node.
>
> If no host is available hte node gets 0 points and "multiplier" points for
> every reachable host.
>
> pacemaker will place your resource on the host that gathers most points. If
> the contact is lost (0 points) the resource moves away. If the contact is
> established again, it depends on the resource-stickiness, if the reource fails
> back to the node.
>
> BTW: you should use ocf:pacemaker:ping instead of the ocf:pacemaker:pingd
> Be sure to have a monitor operation configured.
Thanks for the answer, but I'm too noob and I don't quite follow you :)
I think I understand how the resources are moved according to the pingd 
score. What I need to to is to run a script when the pingd score changes 
from 0 (node disconnected from the network) to gt 0 (node connected to 
the network). This script will resync a database, so it's not a resource 
(AFAIK) and it must run on the newly connected node (not on the node 
that actually has the resource ipaddr2

>
>
>> I tried ClusterMon but it seems to work only when the clone resource is
>> started/stopped, not when the atrtibute value changes
>>
>> Is there a way to have this done ?
>> As last resort I could use a cronjob and monitor the pingd attribute
>> with crm_attribute (crm_attribute --query --name=pingd --node node1
>> returns "name=pingd value=(null) Error performing operation: cib object
>> missing", any idea ?)
>>
>> Any help will be greatly appreciated
>>
>> TIA
>>
>> giuliano
>>
>> versions details
>> OS is centos 5
>> kernel is 2.6.18-194.26.1.el5 #1 SMP Tue Nov 9 12:54:40 EST 2010 i686
>> i686 i386 GNU/Linux
>> heartbeat-3.0.3-2.3.el5.i386.rpm
>> heartbeat-libs-3.0.3-2.3.el5.i386.rpm
>> pacemaker-1.0.12-1.el5.centos.i386.rpm
>> pacemaker-libs-1.0.12-1.el5.centos.i386.rpm
>> all from clusterlabs repo
>>
>>
>> crm config
>> primitive p_ping ocf:pacemaker:ping \
>>           params host_list="10.0.0.254" multiplier="100" dampen="5s" \
>>           op monitor interval="60" timeout="60" \
>>           op start interval="0" timeout="60" \
>>           op stop interval="0" timeout="60"
>
>> clone c_ping p_ping
>> primitive res_IPaddr2 ocf:heartbeat:IPaddr2 \
>>           params ip="10.0.0.80" cidr_netmask="32" \
>>           operations $id="res_IPaddr2-operations" \
>>           op start interval="0" timeout="20" \
>>           op stop interval="0" timeout="20" \
>>           op monitor interval="10" timeout="20" start-delay="0"
>> location l_ping res_IPaddr2 \
>>           rule $id="l_ping-rule" -inf: not_defined pingd or pingd lte 0
> Better:
>
> location l_ping res_IPaddr2 rule pingd: defined pingd
thanks, I'll change that
>> property $id="cib-bootstrap-options" \
>>           dc-version="1.0.12-unknown" \
>>           cluster-infrastructure="Heartbeat" \
>>           stonith-enabled="false" \
>>           no-quorum-policy="ignore" \
>>           last-lrm-refresh="1393595832"
>> rsc_defaults $id="rsc-options" \
>>           resource-stickiness="1000"
> Well, you have a resource stickiness of "1000". A node only can gather 100
> points from the ping resource. if both nodes have 100 points from ping, the
> resource will not fail back. If you want to have fail-back lower the resource-
> stickiness to 50. But why do you want a failback anyway?
in my case I don't want fallback :)

thanks

Giuliano




More information about the Pacemaker mailing list