Re: [Condor-users] NEGOTIATE_ALL_JOBS_IN_CLUSTER

[Steffen Grunewald <steffen.grunewald@xxxxxxxxxx>]

On Wed, Jan 28, 2009 at 12:01:53PM -0600, Todd Tannenbaum wrote:
> 
> Correct, I would keep NEGOTIATE_ALL_JOBS_IN_CLUSTER = False (the 
> default) with Condor v7.0.x or above.
> 
> The problems you mentioned earlier, such as jobs in a cluster not being 
> scheduled if an earlier job in the cluster cannot, should not happen w/ 
> v7.0.x as all jobs are considered "independently" by the autoclustering. 
>   In other words, jobs are "clustered automatically" now by the schedd 
> instead of explicitly by the user - so for negotiation it doesn't matter 
> if you submitted them all in one big cluster or not.

Thanks for the explanation. I'm sure it's in the manual somewhere (which to
be honest sometimes reads like a textbook while the accompaying workbook is
missing). 
I also noticed the new notion of "piespins" - still investigating whether
that would influence the user-prio based job scheduling (compared to 6.x).

> >>> Is it possible to - e.g. once per given time period (4 hours?) - "flush"
> >>> the queue by temporarily setting the macro to True?
> >>>
> >> Maybe something else is going on?  With Condor v7.0.x and above with the 
> >> default auto-clustering, I assert you should never have to resort to 
> >> NEGOTIATE_ALL_JOBS_IN_CLUSTER = True.    Are you over-riding 
> >> autoclustering in your config file by expliciting setting 
> >> SIGNIFICANT_ATTRIBUTES or some such on your condor_config on your submit 
> >> hosts?
> > No, and I never considered to.
> Excellent.  :).

BTW, SIGNIFICANT_ATTRIBUTES only shows up in the text without a separate own
paragraph in the config file entries section. Things have improved a lot,
compared to 6.x manuals... so I'm optimistic in this respect - but to me it's
still hard to identify all the changes being made. It's a bit tricky to find
a quiet place to diff-read two 900+-page manuals :) I had previously suggested
to make updates available as a series of diffs of the original doc code (which
I suppose is LaTEX or SGML or the like), it's more comprehensive than a single
sentence in the Release Notes section. - Any opinions?

Regards,
 Steffen

-- 
Steffen Grunewald * MPI Grav.Phys.(AEI) * Am M�erg 1, D-14476 Potsdam
Cluster Admin * http://pandora.aei.mpg.de/merlin/ * http://www.aei.mpg.de/
* e-mail: steffen.grunewald(*)aei.mpg.de * +49-331-567-{fon:7233,fax:7298}
No Word/PPT mails - http://www.gnu.org/philosophy/no-word-attachments.html