Hi Andrew,
It has just happenned yet again, as soon as I restarted everything. Again
it was your monitoring that actually was the submission that it occurred
on, but I am not convinced that this is not just a co-incidence. Last time
it was just before Liverpool, this time it was (I think because things
mangled) that it was Oxford so no real pattern.
I am going to try cleaning everything in case it has some memory of this
badly formed job_id that keeps re-occurring.
All the best,
david
On Thu, 5 Jun 2003, Andrew Mcnab wrote:
> On Thu, 5 Jun 2003, Dr D J Colling wrote:
>
> > The postgresql database has again become corrupted. Now it looks like the
> > usual problem, however I am getting suspicious. Is it possible for
> > something else to cause this (this is a question to my WP 1 friends)? For
> > example some badly formed jdl with quotation marks?
> >
> > The last five times have been whilst the monitoring jobs that Andy MacNab
> > runs have been submitting. Have you changed anything in the last
> > couple of days Andy?
>
> No changes since last Monday (26th May.) We could try increasing the gap
> between the submissions (it's normally a couple seconds at the moment,
> since it registers the submission with the GridPP webserver after each
> one.) Another thing is that I am trying to submit to all sites, so if
> there's something about one (or more?) of the sites which provokes this,
> then the map job will expose that too. Is there any pattern to which
> recipient site's are associated with it?
>
> I've increased the job submission gap to 60 seconds, so that might fix it
> and will make it easier to look for associations.
>
> Cheers,
>
> Andrew
>
|