I know you asked about clusters and stuff but there is also a
BDII sanity checker in s-2.sf.net for the storage testing. Mind you,
it seems to report some errors which aren't, sort of. But it's
pretty good. Let's cover that on Wed,
--jens
-----Original Message-----
From: Testbed Support for GridPP member institutes on behalf of Steve Traylen
Sent: Mon 23/11/2009 19:20
To: [log in to unmask]
Subject: Re: Site Information Sanity Checking
On Mon, Nov 23, 2009 at 8:14 PM, Matt Doidge <[log in to unmask]> wrote:
> Heya all,
>
> If a site, by whatever mechanism or whatever reason, has a fiddle with their
> information system how can you then check that the info you're pumping out
> tmakes sense without waiting for gstat to refresh? Or to rephrase the
> question, how can we duplicate gstat's giis sanity checking? Maybe some kind
> of magical ldap query with some arcane parsing and grepping?
ldap queries are all ways going to be the true test especially if you
suspect a probe bug.
All the gstat probes are just nagios probes that can be run from the
command line client
or rerun from the Oxford nagios via the web interface. You should have
permission to
do this for your site.
If you install gstat-validate package from the sa1 repository you can
just do things like.
$ gstat-validate-ce -H prod-bdii.cern.ch -p 2170 -b
'Mds-vo-name=CERN-PROD,o=Grid'
there are few other options , see -h for help.
Steve
>
> The backstory of this query is that Lancaster spent a good chunk of Friday
> with the "info" status in gstat as I tried to publish our new subcluster.
> ldapsearch showed that we were pumping out stuff that looked right, but
> gstat complained because the software version wasn't being published for the
> cluster-it appeared to be being pushed out to the world, but not in the
> right place in the publishing apparently. It being Friday we rolled back
> rather then extensively debug, the fact that gstat with its hour-or-so
> between refreshes was our main indicator of our state of play didn't help
> matters. Before I have another go at it I'd like to be armed with a better
> debugging tool.
>
> Thanks in advance,
> Matt
>
--
Steve Traylen
--
Scanned by iCritical.
|