[prev in list] [next in list] [prev in thread] [next in thread] 

List:       linux-ha-dev
Subject:    Re: [Linux-ha-dev] Patch for db2 Resource agent
From:       Dejan Muhamedagic <dejanmm () fastmail ! fm>
Date:       2010-09-16 17:38:46
Message-ID: 20100916173845.GC6163 () rondo ! homenet
[Download RAW message or body]

On Thu, Sep 16, 2010 at 05:18:10PM +0200, Holger Teutsch wrote:
> -----Ursprüngliche Nachricht-----
> Von: Dejan Muhamedagic <dejanmm@fastmail.fm>
> Gesendet: Sep 16, 2010 4:30:37 PM
> An: High-Availability Linux Development List <linux-ha-dev@lists.linux-ha.org>
> Betreff: Re: [Linux-ha-dev] Patch for db2 Resource agent
> 
> > Hi,
> > 
> > On Thu, Sep 16, 2010 at 03:32:43PM +0200, Holger Teutsch wrote:
> > > -----Ursprüngliche Nachricht-----
> > > 
> > > > On Thu, Sep 16, 2010 at 10:07:49AM +0200, Holger Teutsch wrote:
> > > > > Hi,
> > > > > find attached a patch for the db2 resource agent
> > > > > - holger
> > > > > 
> > > > > High: Replace call to db2_local_ps with db2nps (Thanx to Evgeny Nifontov)
> > > > > 
> > > > > db2nodes.cfg typically contains a cluster service address and
> > > > > db2_local_ps never works
> > > > > in this case.
> > > > 
> > > > Is there any documentation or source to support this? Looking at
> > > > the source, db2_local_ps should recover the right node number
> > > > from db2nodes.cfg in case that file exists, then invoke db2nps
> > > > with that node number. If it doesn't work, then it is a bug in
> > > > db2_local_ps.
> > > > 
> > > > Cheers,
> > > > 
> > > > Dejan
> > > 
> > > Dejan,
> > > the background is:
> > > 
> > > db2nodes.cfg maps "partition numbers" to "node names"
> > > Normally this is the physical hostname.
> > > 
> > > Most db2 utilities (e.g. db2start etc) map the node number to the "node name".
> > > If "node name" == "uname -n" call locally, else invoke the
> > > command on the remote host via rsh or something you set via
> > > 
> > > db2set DB2RSHCMD=xxx
> > > .
> > > 
> > > On a cluster the file ~/sqllib/db2nodes.cfg is on a shared FS.
> > > 
> > > If this file contains a physical hostname this will definitely be wrong after a \
> > > resource migration. 
> > > Therefore per best practice (IBM Redbooks etc.) you put the DNS
> > > name of the cluster service into db2nodes.cfg.
> > > 
> > > You then have to take care that the instance owner has rsh access to itself via \
> > > the service address. Else you can map  DB2RSHCMD to a script like
> > > 
> > > ---
> > > #!/bin/sh
> > > #
> > > # Emulate inter partition call for ACC by simply do it locally
> > > # Install with
> > > #
> > > #   db2set DB2RSHCMD=$INSTHOME/db2_acc_rsh
> > > #
> > > # HT Dec 2007
> > > 
> > > # Is called rsh like as
> > > # c36db.hg.fresenius.de -n -l db2c36 ARGS
> > > 
> > > # remove 4 first args
> > > shift
> > > shift
> > > shift
> > > shift
> > > eval "$@"
> > > ----
> > > 
> > > Other methods for working with db2nodes.cfg are
> > > - make links to node local files (problems with permissions)
> > > - edit db2nodes.cfg with the resource agent prior to an (something a IBM povide \
> > > lib does/tries) 
> > > You see we are veterans with 100+ db2 databases. 8-) or 8-(
> > > 
> > > I agree that the logic of db2_local_ps is to simple to deal with this setup.
> > > On the other hand I'm not confident that IBM will implement a comprehensive way \
> > > to detect "locality" for a general multi partition setup. This would require \
> > > checking IP interfaces etc... 
> > > I checked that db2nps works on DB2 8.2, 9.5 and 9.7.
> > > 
> > > Any concerns let me know.
> > 
> > My only concern is that 0 may not be the right node number for
> > the local node. On failover or failback or whenever. That's why
> > I'd expect db2_local_ps to be the right way to go. Of course, I
> > do understand that things don't always work as advertised and in
> > particular with products such as db2 :) At any rate, if you're
> > sure that "db2nps 0" is always going to refer to the local node,
> > then we can go ahead with this change.
> > 
> > Cheers,
> > 
> > Dejan
> > 
> 
> Dejan,
> in a single partition database the single partition is always "0".
> 
> db2nps NN   looks on the *local* node for processes of partition "NN"
> 
> i.e. db2nps 0 does the required work.
> 
> So I think that's OK.

Fine. I pushed the patch. Many thanks for the contribution and
taking time to explain this.

Cheers,

Dejan

> Regards
> Holger
> ___________________________________________________________
> GRATIS: Spider-Man 1-3 sowie 300 weitere Videos!
> Jetzt kostenlose Movie-FLAT freischalten! http://movieflat.web.de
> _______________________________________________________
> Linux-HA-Dev: Linux-HA-Dev@lists.linux-ha.org
> http://lists.linux-ha.org/mailman/listinfo/linux-ha-dev
> Home Page: http://linux-ha.org/
_______________________________________________________
Linux-HA-Dev: Linux-HA-Dev@lists.linux-ha.org
http://lists.linux-ha.org/mailman/listinfo/linux-ha-dev
Home Page: http://linux-ha.org/


[prev in list] [next in list] [prev in thread] [next in thread]