Issues With HeartBeat 2.0.8 After reboot of servers on a 2 Node Cluster

I am working on HA 2.0.8 with a 2 Node cluster. We have 2
resources for
which I have set the scores for both the nodes in the
cib.xml

I am pasting you the part of the code where I set the
scores. The switch
over happens fine as expected upon a stop of the resource on
any node. When
I reboot a node, the other node gets the resource to run,
but when the first
node is back UP, the resource is forcefully migrated back to
the node which
just rebooted.

I think this happens because of the following(Here Active
means that the
node is running the resources and Standby Means that the
node is not running
the resources) :

1. Node 1 (Active , score 5000); Node 2 (Standby , Score
4995)

2. Switchover =>
Node 1 (standby, score 4990), Node 2 (Active, Score 4995).

3. reboot 1 =>
Node 1 -> Heartbeat restarts, fail count reset =>
Score on node 1 = 5000 and
Node 2 which gets to run the resource while the Node 1 was
rebooting still
has the score of 4995.(Less than 5000).

Please let me know a way in which I can override the scores
and tell the
heartbeat not to look into the scores in the case a node is
running a
resource and a new node comes into the cluster with a higher
score for the
same resource.

If you think I can work this out with the scores too, please
let me know; it
would be even better 

Please let me know.

Thanks in Advance!

                <constraints>
                        <rsc_location
id="rsc_location_group_1"
rsc="group_1">
                                <rule
id="prefered_location_group_1"
score="5000">
                                        <expression
attribute="#uname"
id="prefered_location_group_1_expr"
operation="eq" value="ACTIVEHOST"/>
                                </rule>
                                <rule
id="prefered_location_group_2"
score="4995">
                                        <expression
attribute="#uname"
id="prefered_location_group_2_expr"
operation="eq" value="STANDBYHOST"/>
                                </rule>
                        </rsc_location>
                </constraints>
_______________________________________________
Linux-HA mailing list
Linux-HAlists.linux-ha.org
h
ttp://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha
.org/ReportingProblems

Adding to the point 3 below -> Since the score of Node 1
after a reboot is
5000, more than that of the other node, it seems to force
heartbeat to
switchover a running resource on node 2.

We want to tell heartbeat not to do that because upon a
reboot, Node 1
should silently come and join as a standby Node and wait to
run the resource
when Node 2 (the current Active) fails to run it.

So basically, we want scores to come into picture only in
the case of
failure and not in the case of a new node coming into the
cluster with a
higher score.

Thanks again
Raghu

On Thu, Jul 24, 2008 at 12:23 AM, raghu ram
<p.raghuramgmail.com> wrote:

> I am working on HA 2.0.8 with a 2 Node cluster. We have
2 resources for
> which I have set the scores for both the nodes in the
cib.xml
>
> I am pasting you the part of the code where I set the
scores. The switch
> over happens fine as expected upon a stop of the
resource on any node. When
> I reboot a node, the other node gets the resource to
run, but when the first
> node is back UP, the resource is forcefully migrated
back to the node which
> just rebooted.
>
> I think this happens because of the following(Here
Active means that the
> node is running the resources and Standby Means that
the node is not running
> the resources) :
>
> 1. Node 1 (Active , score 5000); Node 2 (Standby ,
Score 4995)
>
> 2. Switchover =>
> Node 1 (standby, score 4990), Node 2 (Active, Score
4995).
>
> 3. reboot 1 =>
> Node 1 -> Heartbeat restarts, fail count reset =>
Score on node 1 = 5000
> and Node 2 which gets to run the resource while the
Node 1 was rebooting
> still has the score of 4995.(Less than 5000).
>
> Please let me know a way in which I can override the
scores and tell the
> heartbeat not to look into the scores in the case a
node is running a
> resource and a new node comes into the cluster with a
higher score for the
> same resource.
>
> If you think I can work this out with the scores too,
please let me know;
> it would be even better 
>
> Please let me know.
>
> Thanks in Advance!
>
>                 <constraints>
>                         <rsc_location
id="rsc_location_group_1"
> rsc="group_1">
>                                 <rule
id="prefered_location_group_1"
> score="5000">
>                                         <expression
attribute="#uname"
> id="prefered_location_group_1_expr"
operation="eq" value="ACTIVEHOST"/>
>                                 </rule>
>                                 <rule
id="prefered_location_group_2"
> score="4995">
>                                         <expression
attribute="#uname"
> id="prefered_location_group_2_expr"
operation="eq" value="STANDBYHOST"/>
>                                 </rule>
>                         </rsc_location>
>                 </constraints>
>
>
>
_______________________________________________
Linux-HA mailing list
Linux-HAlists.linux-ha.org
h
ttp://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha
.org/ReportingProblems

On Thu, Jul 24, 2008 at 09:29, raghu ram <p.raghuramgmail.com> wrote:
> Adding to the point 3 below -> Since the score of
Node 1 after a reboot is
> 5000, more than that of the other node, it seems to
force heartbeat to
> switchover a running resource on node 2.
>
> We want to tell heartbeat not to do that because upon a
reboot, Node 1
> should silently come and join as a standby Node and
wait to run the resource
> when Node 2 (the current Active) fails to run it.
>
> So basically, we want scores to come into picture only
in the case of
> failure and not in the case of a new node coming into
the cluster with a
> higher score.

no.  thats not how the cluster works.
scores are the _only_ thing that determines where a resource
runs.

try looking up default-resource-stickiness

>
> Thanks again
> Raghu
>
> On Thu, Jul 24, 2008 at 12:23 AM, raghu ram
<p.raghuramgmail.com> wrote:
>
>> I am working on HA 2.0.8 with a 2 Node cluster. We
have 2 resources for
>> which I have set the scores for both the nodes in
the cib.xml
>>
>> I am pasting you the part of the code where I set
the scores. The switch
>> over happens fine as expected upon a stop of the
resource on any node. When
>> I reboot a node, the other node gets the resource
to run, but when the first
>> node is back UP, the resource is forcefully
migrated back to the node which
>> just rebooted.
>>
>> I think this happens because of the following(Here
Active means that the
>> node is running the resources and Standby Means
that the node is not running
>> the resources) :
>>
>> 1. Node 1 (Active , score 5000); Node 2 (Standby ,
Score 4995)
>>
>> 2. Switchover =>
>> Node 1 (standby, score 4990), Node 2 (Active, Score
4995).
>>
>> 3. reboot 1 =>
>> Node 1 -> Heartbeat restarts, fail count reset
=> Score on node 1 = 5000
>> and Node 2 which gets to run the resource while the
Node 1 was rebooting
>> still has the score of 4995.(Less than 5000).
>>
>> Please let me know a way in which I can override
the scores and tell the
>> heartbeat not to look into the scores in the case a
node is running a
>> resource and a new node comes into the cluster with
a higher score for the
>> same resource.
>>
>> If you think I can work this out with the scores
too, please let me know;
>> it would be even better 
>>
>> Please let me know.
>>
>> Thanks in Advance!
>>
>>                 <constraints>
>>                         <rsc_location
id="rsc_location_group_1"
>> rsc="group_1">
>>                                 <rule
id="prefered_location_group_1"
>> score="5000">
>>                                        
<expression attribute="#uname"
>> id="prefered_location_group_1_expr"
operation="eq" value="ACTIVEHOST"/>
>>                                 </rule>
>>                                 <rule
id="prefered_location_group_2"
>> score="4995">
>>                                        
<expression attribute="#uname"
>> id="prefered_location_group_2_expr"
operation="eq" value="STANDBYHOST"/>
>>                                 </rule>
>>                         </rsc_location>
>>                 </constraints>
>>
>>
>>
> _______________________________________________
> Linux-HA mailing list
> Linux-HAlists.linux-ha.org
> h
ttp://lists.linux-ha.org/mailman/listinfo/linux-ha
> See also: http://linux-ha
.org/ReportingProblems
>
_______________________________________________
Linux-HA mailing list
Linux-HAlists.linux-ha.org
h
ttp://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha
.org/ReportingProblems