[gpfsug-discuss] Forcing which node gets expelled?

[Felipe Knop]

All,

As Bob Oesterlin indicated, it is possible to define an expel script (see 
/usr/lpp/mmfs/samples/expelnode.sample) to control which of the two nodes 
to get expelled. The script can also be used to issue alerts, etc.

The current policy used (before the script is invoked) when deciding which 
node to expel is the following:

       1. quorum nodes over non-quorum nodes

       2. local nodes over remote nodes

       3. manager-capable nodes over non-manager-capable nodes

       4. nodes managing more FSs over nodes managing fewer FSs

       5. NSD server over non-NSD server

       Otherwise, expel whoever joined the cluster more recently.



The statement below from Dr. Uwe Falke is also correct: addressing the 
network connectivity is the better long-term approach, but the callback 
script could be used to control which node to expel.

  Felipe

----
Felipe Knop                                     knop at us.ibm.com
GPFS Development and Security
IBM Systems
IBM Building 008
2455 South Rd, Poughkeepsie, NY 12601
(845) 433-9314  T/L 293-9314





From:   "Uwe Falke" <UWEFALKE at de.ibm.com>
To:     gpfsug main discussion list <gpfsug-discuss at spectrumscale.org>
Date:   10/25/2016 08:32 AM
Subject:        Re: [gpfsug-discuss] Forcing which node gets expelled?
Sent by:        gpfsug-discuss-bounces at spectrumscale.org



Usually, the cluster mgr, receiving a complaint from a node about another 
node being gone, checks its own connection to that other node. If that is 
positive it expells the requester, if not it follows the request and 
expells the other node. 
AFAIK, there are some more subtle algorithms in place if managers or 
quorum nodes are affected. Maybe that can be used to protect certain nodes 

from getting expelled by assigning some role in the cluster to them. I do 
however not know these exactly.

That means: it is not easily controllable which one gets expelled. 
It is better to concentrate on fixing your connectivity issues, as GPFS 
will not feel comfortable in such a unreliable environment anyway. 
 
Mit freundlichen Grüßen / Kind regards

 
Dr. Uwe Falke
 
IT Specialist
High Performance Computing Services / Integrated Technology Services / 
Data Center Services
-------------------------------------------------------------------------------------------------------------------------------------------
IBM Deutschland
Rathausstr. 7
09111 Chemnitz
Phone: +49 371 6978 2165
Mobile: +49 175 575 2877
E-Mail: uwefalke at de.ibm.com
-------------------------------------------------------------------------------------------------------------------------------------------
IBM Deutschland Business & Technology Services GmbH / Geschäftsführung: 
Frank Hammer, Thorsten Moehring
Sitz der Gesellschaft: Ehningen / Registergericht: Amtsgericht Stuttgart, 
HRB 17122 




From:   Matt Thorpe <matt.thorpe at bodleian.ox.ac.uk>
To:     gpfsug main discussion list <gpfsug-discuss at spectrumscale.org>
Date:   10/25/2016 02:05 PM
Subject:        [gpfsug-discuss] Forcing which node gets expelled?
Sent by:        gpfsug-discuss-bounces at spectrumscale.org



Hi,

We are in the process of diagnosing a networking issue that is causing 2 
of our 6 node GPFS cluster to expel each other (it appears they experience 

a temporary network connection outage and lose contact with each other). 
At present it's not consistent which gets expelled by the cluster manager, 

and I wondered if there was any way to always force a specific node to be 
expelled in this situation?

Thanks and best regards,
Matt

--------
Matt Thorpe  |  BDLSS Systems Administrator
Bodleian Libraries Osney One Building, Osney Mead, Oxford, OX2 0EW
matt.thorpe at bodleian.ox.ac.uk  |  01865 (2)80027
_______________________________________________
gpfsug-discuss mailing list
gpfsug-discuss at spectrumscale.org
http://gpfsug.org/mailman/listinfo/gpfsug-discuss





_______________________________________________
gpfsug-discuss mailing list
gpfsug-discuss at spectrumscale.org
http://gpfsug.org/mailman/listinfo/gpfsug-discuss




-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://gpfsug.org/pipermail/gpfsug-discuss_gpfsug.org/attachments/20161025/3dd11db6/attachment-0002.htm>