>>>
From: Jean-Francois Malouin<[EMAIL PROTECTED]>
To:linux-ha@lists.linux-ha.org
Date: 30/07/08 20:39
Subject: Re: [Linux-HA] resources don't migrate when node is declared dead?!?
* ZiLioN ZilLioN <[EMAIL PROTECTED]> [20080729 19:03]:
>
>
>
> > Date: Tue, 29 Jul 2008 17:27:38 -0400
> > From: [EMAIL PROTECTED]
> > To: linux-ha@lists.linux-ha.org
> > Subject: [Linux-HA] resources don't migrate when node is declared dead?!?
> >
> > My cluster contains 2 active/passive nodes with one drbd master/slave
> > resource and one group resource which itself contains 7 resources. I
> > want the m/s and group to be colocated and when the master loose it's
> > ping then the slave should be promoted but nothing happens when I
> > pulled the ethernet cable... Here's what the constrains look like in
> > the cib right now:
> >
[...]
>
> > <rsc_location id="drbd_id:connected" rsc="ms-drbd_id">
> > <rule role="master" id="drbd_id:connected:rule"
> > score_attribute="pingd">
> > <expression id="drbd_id:connected-rule-1" attribute="pingd"
> > operation="defined"/>
> > </rule>
>
> If your score_attribute="pingd", your pingd scaling factor is 100, then
> having access to one node is worth 100, 2 nodes is worth 200, and so on.
> If you don´t have connectivity score_attribute=0.
>
>
> > </rsc_location>
> > <rsc_location id="cli-prefer-mysql_id" rsc="mysql_id">
> > <rule id="cli-prefer-rule-mysql_id" score="INFINITY">
> > <expression id="cli-prefer-expr-mysql_id" attribute="#uname"
> > operation="eq" value="feeble-1" type="string"/>
> > </rule>
>
> If the group_id runs on the same machine as ms-drbd_id, you can not specify
> to mysql_id run always in feeble-1, because mysql_id belongs to the group
> group_id. Your group_id runs on the same machine as ms-drbd_id and ms-drbd_id
> runs in the node with the best connectivity.
>
> > </rsc_location>
> > <rsc_location id="cli-prefer-drbd_id:0" rsc="drbd_id:0">
> > <rule id="cli-prefer-rule-drbd_id:0" score="INFINITY">
> > <expression id="cli-prefer-expr-drbd_id:0"
> > attribute="#uname" operation="eq" value="feeble-0" type="string"/>
> > </rule>
> > </rsc_location>
> > <rsc_location id="cli-prefer-drbd_id:1" rsc="drbd_id:1">
> > <rule id="cli-prefer-rule-drbd_id:1" score="INFINITY">
> > <expression id="cli-prefer-expr-drbd_id:1"
> > attribute="#uname" operation="eq" value="feeble-0" type="string"/>
> > </rule>
> > </rsc_location>
> > </constraints>
I have no clue where these constraints come from. I certainly didn't
put them manually...maybe the crm did while I was messing things up.
In any case I have removed those and replace the pingd constraint
as on the doc on ha-linux suggests:
I believe that these are there because you did a migrate resource
<rsc_location id="ms-drbd_id:connected" rsc="ms-drbd_id">
<rule id="ms-drbd_id:connected:rule" score="-INFINITY" boolean_op="or">
<expression id="ms-drbd_id:connected:expr:undefined" attribute="pingd"
operation="not_defined"/>
<expression id="ms-drbd_id:connected:expr:zero" attribute="pingd"
operation="lte" value="0"/>
</rule>
</rsc_location>
and when I pull the cable on the master all the resources migrate to
the other nope. I'm happy with that but I face another problem and
I'll start a new thread on it rather than continuing here.
> > ~# showscores.sh
> > Resource Score Node Stickiness #Fail
> > Fail-Stickiness
> > -1000000_(master) -INFINITY ptest[11262]: 100 -1001
> >
> > 1000000_(master) INFINITY ptest[11262]: 100 -1001
> > 100_(master) 100 ptest[11262]: 100 -1001
> > 175_(master) 175 ptest[11262]: 100 -1001
> > 76_(master) 76 ptest[11262]: 100 -1001
>
> I don´t understand this, where is the name of resources? node?
I updated the showscores script as found on
http://hg.clusterlabs.org/pacemaker/dev/file/tip/contrib/showscores.sh
and now my scores looks like:
Resource Score Node Stickiness #Fail Fail-Stickiness
apache_id 0 feeble-0 100 0 -1001
apache_id 600 feeble-1 100 0 -1001
drbd_id:0 -INFINITY feeble-0 100 0 -1001
drbd_id:0 75 feeble-1 100 0 -1001
drbd_id:0_(master) INFINITY feeble-1 100 0 -1001
drbd_id:1 76 feeble-1 100 0 -1001
drbd_id:1 INFINITY feeble-0 100 0 -1001
drbd_id:1_(master) 75 feeble-0 100 0 -1001
drbd_id:2 -INFINITY feeble-0 100 0 -1001
drbd_id:2 -INFINITY feeble-1 100 0 -1001
fs_id 0 feeble-0 100 0 -1001
fs_id 600 feeble-1 100 0 -1001
group_id 0 feeble-0 100 0 -1001
group_id 600 feeble-1 100 0 -1001
ip_id 0 feeble-0 100 0 -1001
ip_id 600 feeble-1 100 0 -1001
mysql_id 0 feeble-0 100 0 -1001
mysql_id INFINITY feeble-1 100 0 -1001
nfs_common-id 0 feeble-0 100 0 -1001
nfs_common-id 600 feeble-1 100 0 -1001
nfs_kernel-id 0 feeble-0 100 0 -1001
nfs_kernel-id 600 feeble-1 100 0 -1001
I do not understand yet how exactly the scores are calculated for a
master-slave resource and I don't know if the showscores script deals
with them correctly...
>
> >
> > Any ideas?
>
> Before pulled the ethernet cable, try:
>
> cibadmin -Q -o nodes
>
> You can see the all the nodes of your cluster?
> Today in my cluster only a node of the two was recognized.
both nodes are there.
>
>
> I do not think you have helped a lot, even I do not have much
> experience with the cib.xml. I hope that the comments at least you are
> beneficial.
thanks anyway.
jf
--
<° ><
_______________________________________________
Linux-HA mailing list
Linux-HA@lists.linux-ha.org
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems
_______________________________________________
Linux-HA mailing list
Linux-HA@lists.linux-ha.org
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems
