Login | Register For Free | Help
Search for: (Advanced)

Mailing List Archive: Linux-HA: Pacemaker

the behavior of resource-stickiness

 

 

Linux-HA pacemaker RSS feed   Index | Next | Previous | View Threaded


ikedaj at intellilink

Nov 19, 2009, 4:22 AM

Post #1 of 4 (382 views)
Permalink
the behavior of resource-stickiness

Hi,

I am running pacemaker 1.0.6 + corosync 1.1.2.
There are two resource and their "resource-stickiness" is "INFINITY".
They have rsc_colocation but there is no rsc_location setting.

(1) start the resources first
# crm_mon -1

============
Last updated: Thu Nov 19 20:35:45 2009
Stack: openais
Current DC: dom1 - partition with quorum
Version: 1.0.6-f709c638237cdff7556cb6ab615f32826c0f8c06
3 Nodes configured, 2 expected votes
2 Resources configured.
============

Online: [ dom9 dom1 dom2 ]

dummy1 (ocf::heartbeat:Dummy): Started dom1
dummy2 (ocf::heartbeat:Dummy): Started dom9


(2) standby one node
# crm_standby -U dom1 -v on

I expected that dummy2 kept running on dom9,
but it moved to another node.

# crm_mon -1
============
Last updated: Thu Nov 19 20:36:27 2009
Stack: openais
Current DC: dom1 - partition with quorum
Version: 1.0.6-f709c638237cdff7556cb6ab615f32826c0f8c06
3 Nodes configured, 2 expected votes
2 Resources configured.
============

Node dom1: standby
Online: [ dom9 dom2 ]

dummy1 (ocf::heartbeat:Dummy): Started dom9
dummy2 (ocf::heartbeat:Dummy): Started dom2


(2) standby "off" one node
# crm_standby -U dom1 -v off

Both dummy1 and dummy2 moved to the other node.
It's strange because these resources' "resource-stickiness" is "INFINITY".
Is there anything wrong in my setting?


# crm_mon -1
============
Last updated: Thu Nov 19 20:36:50 2009
Stack: openais
Current DC: dom1 - partition with quorum
Version: 1.0.6-f709c638237cdff7556cb6ab615f32826c0f8c06
3 Nodes configured, 2 expected votes
2 Resources configured.
============

Online: [ dom9 dom1 dom2 ]

dummy1 (ocf::heartbeat:Dummy): Started dom2
dummy2 (ocf::heartbeat:Dummy): Started dom1




I attached the log.

Thanks,
Junko
Attachments: standby.tar.gz (48.4 KB)


andrew at beekhof

Nov 19, 2009, 11:41 AM

Post #2 of 4 (364 views)
Permalink
Re: the behavior of resource-stickiness [In reply to]

2009/11/19 Junko IKEDA <ikedaj [at] intellilink>:
> Hi,
>
> I am running pacemaker 1.0.6 + corosync 1.1.2.
> There are two resource and their "resource-stickiness" is "INFINITY".
> They have rsc_colocation but there is no rsc_location setting.
>
> (1) start the resources first
> # crm_mon -1
>
> ============
> Last updated: Thu Nov 19 20:35:45 2009
> Stack: openais
> Current DC: dom1 - partition with quorum
> Version: 1.0.6-f709c638237cdff7556cb6ab615f32826c0f8c06
> 3 Nodes configured, 2 expected votes
> 2 Resources configured.
> ============
>
> Online: [ dom9 dom1 dom2 ]
>
> dummy1  (ocf::heartbeat:Dummy): Started dom1
> dummy2  (ocf::heartbeat:Dummy): Started dom9
>
>
> (2) standby one node
> # crm_standby -U dom1 -v on
>
> I expected that dummy2 kept running on dom9,
> but it moved to another node.
>
> # crm_mon -1
> ============
> Last updated: Thu Nov 19 20:36:27 2009
> Stack: openais
> Current DC: dom1 - partition with quorum
> Version: 1.0.6-f709c638237cdff7556cb6ab615f32826c0f8c06
> 3 Nodes configured, 2 expected votes
> 2 Resources configured.
> ============
>
> Node dom1: standby
> Online: [ dom9 dom2 ]
>
> dummy1  (ocf::heartbeat:Dummy): Started dom9
> dummy2  (ocf::heartbeat:Dummy): Started dom2
>
>
> (2) standby "off" one node
> # crm_standby -U dom1 -v off
>
> Both dummy1 and dummy2 moved to the other node.
> It's strange because these resources' "resource-stickiness" is "INFINITY".
> Is there anything wrong in my setting?

This wont be helping:

Nov 19 20:36:50 dom1 pengine: [3180]: info: native_merge_weights:
dummy2: Breaking dependancy loop at dummy1
Nov 19 20:36:50 dom1 pengine: [3180]: info: native_merge_weights:
dummy1: Breaking dependancy loop at dummy2

You really don't want to create colocation loops, they'll cause the
cluster to behave somewhat erratically (as you're seeing).
I removed the following constraint and the resources stopped moving:
<rsc_colocation id="location2" rsc="dummy2" with-rsc="dummy1"
score="-INFINITY"/>


> # crm_mon -1
> ============
> Last updated: Thu Nov 19 20:36:50 2009
> Stack: openais
> Current DC: dom1 - partition with quorum
> Version: 1.0.6-f709c638237cdff7556cb6ab615f32826c0f8c06
> 3 Nodes configured, 2 expected votes
> 2 Resources configured.
> ============
>
> Online: [ dom9 dom1 dom2 ]
>
> dummy1  (ocf::heartbeat:Dummy): Started dom2
> dummy2  (ocf::heartbeat:Dummy): Started dom1
>
>
>
>
> I attached the log.
>
> Thanks,
> Junko
>
> _______________________________________________
> Pacemaker mailing list
> Pacemaker [at] oss
> http://oss.clusterlabs.org/mailman/listinfo/pacemaker
>
>

_______________________________________________
Pacemaker mailing list
Pacemaker [at] oss
http://oss.clusterlabs.org/mailman/listinfo/pacemaker


andrew at beekhof

Nov 19, 2009, 11:41 AM

Post #3 of 4 (363 views)
Permalink
Re: the behavior of resource-stickiness [In reply to]

2009/11/19 Junko IKEDA <ikedaj [at] intellilink>:
> Hi,
>
> I am running pacemaker 1.0.6 + corosync 1.1.2.
> There are two resource and their "resource-stickiness" is "INFINITY".
> They have rsc_colocation but there is no rsc_location setting.
>
> (1) start the resources first
> # crm_mon -1
>
> ============
> Last updated: Thu Nov 19 20:35:45 2009
> Stack: openais
> Current DC: dom1 - partition with quorum
> Version: 1.0.6-f709c638237cdff7556cb6ab615f32826c0f8c06
> 3 Nodes configured, 2 expected votes
> 2 Resources configured.
> ============
>
> Online: [ dom9 dom1 dom2 ]
>
> dummy1  (ocf::heartbeat:Dummy): Started dom1
> dummy2  (ocf::heartbeat:Dummy): Started dom9
>
>
> (2) standby one node
> # crm_standby -U dom1 -v on
>
> I expected that dummy2 kept running on dom9,
> but it moved to another node.
>
> # crm_mon -1
> ============
> Last updated: Thu Nov 19 20:36:27 2009
> Stack: openais
> Current DC: dom1 - partition with quorum
> Version: 1.0.6-f709c638237cdff7556cb6ab615f32826c0f8c06
> 3 Nodes configured, 2 expected votes
> 2 Resources configured.
> ============
>
> Node dom1: standby
> Online: [ dom9 dom2 ]
>
> dummy1  (ocf::heartbeat:Dummy): Started dom9
> dummy2  (ocf::heartbeat:Dummy): Started dom2
>
>
> (2) standby "off" one node
> # crm_standby -U dom1 -v off
>
> Both dummy1 and dummy2 moved to the other node.
> It's strange because these resources' "resource-stickiness" is "INFINITY".
> Is there anything wrong in my setting?

This wont be helping:

Nov 19 20:36:50 dom1 pengine: [3180]: info: native_merge_weights:
dummy2: Breaking dependancy loop at dummy1
Nov 19 20:36:50 dom1 pengine: [3180]: info: native_merge_weights:
dummy1: Breaking dependancy loop at dummy2

You really don't want to create colocation loops, they'll cause the
cluster to behave somewhat erratically (as you're seeing).
I removed the following constraint and the resources stopped moving:
<rsc_colocation id="location2" rsc="dummy2" with-rsc="dummy1"
score="-INFINITY"/>


> # crm_mon -1
> ============
> Last updated: Thu Nov 19 20:36:50 2009
> Stack: openais
> Current DC: dom1 - partition with quorum
> Version: 1.0.6-f709c638237cdff7556cb6ab615f32826c0f8c06
> 3 Nodes configured, 2 expected votes
> 2 Resources configured.
> ============
>
> Online: [ dom9 dom1 dom2 ]
>
> dummy1  (ocf::heartbeat:Dummy): Started dom2
> dummy2  (ocf::heartbeat:Dummy): Started dom1
>
>
>
>
> I attached the log.
>
> Thanks,
> Junko
>
> _______________________________________________
> Pacemaker mailing list
> Pacemaker [at] oss
> http://oss.clusterlabs.org/mailman/listinfo/pacemaker
>
>

_______________________________________________
Pacemaker mailing list
Pacemaker [at] oss
http://oss.clusterlabs.org/mailman/listinfo/pacemaker


ikedaj at intellilink

Nov 19, 2009, 5:51 PM

Post #4 of 4 (357 views)
Permalink
Re: the behavior of resource-stickiness [In reply to]

Hi,

On Fri, 20 Nov 2009 04:41:10 +0900, Andrew Beekhof <andrew [at] beekhof>
wrote:

> You really don't want to create colocation loops, they'll cause the
> cluster to behave somewhat erratically (as you're seeing).
> I removed the following constraint and the resources stopped moving:
> <rsc_colocation id="location2" rsc="dummy2" with-rsc="dummy1"
> score="-INFINITY"/>

Thanks a lot!
I could get the desired outcome.
Thank you for your advice.

Best Regards,
Junko

_______________________________________________
Pacemaker mailing list
Pacemaker [at] oss
http://oss.clusterlabs.org/mailman/listinfo/pacemaker

Linux-HA pacemaker RSS feed   Index | Next | Previous | View Threaded
 
 


Interested in having your list archived? Contact Gossamer Threads
 
  Web Applications & Managed Hosting Powered by Gossamer Threads Inc.