[Linux-HA] IP Power 9258 stonith plugin

Helmut Weymann helmut at h-weymann.de
Mon Mar 8 14:56:01 MST 2010


Hi,

Am Montag, 8. März 2010 schrieb Dejan Muhamedagic:
> Hi,
>
> Sorry for the delay and thanks for the reminder.
>
> On Mon, Mar 08, 2010 at 08:25:05AM +0100, Helmut Weymann wrote:
> > Hi,
> >
> > Am Freitag, 26. Februar 2010 schrieb Dejan Muhamedagic:
> > > On Fri, Feb 26, 2010 at 11:54:41AM +0100, Dejan Muhamedagic wrote:
> > > > Hi,
> > > >
> > > > On Fri, Feb 26, 2010 at 01:23:17AM +0100, Helmut Weymann wrote:
> > >
> > > [...]
> > >
> > > > Well, if that's the only way, let's have it like that. Perhaps
> > > > also to allow for a host-port map file, which would be preferable
> > > > if for instance ports 7 and 8 are used. Though that file could be
> > > > a maintenance issue, with csync2 or rsync it could be kept in
> > > > good shape.
> > > >
> > > > The parameter hostlist should be renamed to port_map or similar.
> > >
> > > Actually not, hostlist is better.
> > >
> > > A few more suggestions:
> > >
> > > - you could allow for "empty" slots other values apart from NULL
> > >   (say "N" or "X")
> >
> > Managed to get a list from the device via the web-interface.
>
> Great!
>
> > Nevertheless I still allow to define the parameter hostlist.
>
> Probably not a bad idea, in case the device is used to manage
> more than just this cluster.
>
> > Names used in that list are only used if CRM requests to kill a
> > cluster-node. (thats my guess.)  So any name should be allowed. Only the
> > names of cluster nodes should make a difference.
>
> It would make sense to also check if the names in the given
> hostlist are available for management at the device.
>
> > Please review the attached submission. It includes:
> > - external/ippower9258 (the plugin in bin folder)
> > - README.ippower9258 in doc folder.
> >
> > As an addon I have included in folder doc/example the relevant html-files
> > the device creates.
>
> Thanks, that was helpful.
>
> > I have marked a few lines of the plugin with the comment "please review".
> > These are about loglevels supported by ha_log.sh, pathnames below
> > /var/run/heartbeat/rsctmp and a piece of code between the lines 233 to
> > 263.
> >
> > Please feel free to make any adjustments. I'll run the necessary tests.
>
> You'll find a patch attached. It contains the following changes:
>
> - different build_hostlist procedure based on output of w3m -dump
> - exit on error in get_*_from_device
> - a few other minor changes
I did apply your patch and tested the new plugin with the stonith command.
The different tests, did what they were expected to do: got the hostlist, 
status was confirmed, port status was under control.
Just when I used a wrong hostname the error message printed each hostname from 
the list on a separate line. - I hope this is just about nice output.
As we no longer need the devicename it is superfluous to get DEVICE_SETTINGS.
I will remove that code including the corresponding error handling.
>
> BTW, if you could leave out most of {} around variable
> references, that'd make my eyes less wobble.
I'm completely new to shell scripting. Do you have any hint on when to use 
these {} bracket pairs? Nevertheless give me a few days to remove the brackets 
and have the outcome tested.
>
> Cheers,
>
> Dejan
>
> > > - take out the documentation part (comments at the top of the
> > >   file) and put it in a separate README file (README.ippower9258)
> >
> > Done
> >
> > > - please make the indentation consistent, sometimes there are
> > >   spaces, sometime tabs
> >
> > Done.
> >
> > > Cheers,
> > >
> > > Dejan
> >
> > Regards,
> >
> > Helmut
> >
> >
> >
> >
> >
> > _______________________________________________
> > Linux-HA mailing list
> > Linux-HA at lists.linux-ha.org
> > http://lists.linux-ha.org/mailman/listinfo/linux-ha
> > See also: http://linux-ha.org/ReportingProblems





More information about the Linux-HA mailing list