[intrepid-notify] Re: clusterbank upgrade

[Jonathon Anderson]

The clusterbank upgrade is complete. Clusterbank 0.4 is now available
on all login machines for Surveyor, Gadzooks, Intrepid, and Eureka.
Job submission should now be working normally. Jobs running on
Gadzooks and Eureka are now required to be related to a specific
project, which might need to be specified using the -A option to qsub.
(Gadzooks users should use their Surveyor project; Eureka users should
user their Intrepid project.)

Clusterbank is no longer being managed by softenv, so you no longer
need to have +clusterbank specified in your environment. Because
clusterbank 0.4 is incompatible with 0.3.x, the versions of
clusterbank that were once available in softenv are no longer
available for use.

0.4 adds the ability to track jobs. As a result, the reports have
replaced the "Charges" column with the "Jobs" column. You can now
count on the value of this column being canonically the number of jobs
run.

A major bug has been fixed regarding the behavior of charges and
refunds applied to expired allocations. The "available" amount of an
allocation, as seen in both "cbank list allocations" and "cbank list
projects" is now 0 for any expired allocation.

What was once the "report" metacommand has become the "list"
metacommand. If you type "cbank-report-<command>" or "cbank report
<command>", now type "cbank-list" or "cbank list" in stead. If you
never specified the "report" keyword explicitly, nothing has changed:
the "list" metacommand is the default, and behaves exactly the same as
"report".

It has been pointed out to me that the man pages for cbank are
difficult to follow, due to the metacommand architecture. (In fact, I
might be the only one that knows what I mean by "metacommand.") As
such, I am starting a standard manual at
http://trac.mcs.anl.gov/projects/clusterbank/wiki/Manual/cbank that
should make things much clearer. As is, it only describes how to find
the appropriate man pages, but I will continue working on it,
including an initial contribution tomorrow.

As always, I welcome and appreciate any feedback on clusterbank and
cbank. If you have problems, questions, or suggestions, email me at
janderso at alcf.anl.gov, or alcf support at support at alcf.anl.gov.

Thank you.
~jonathon anderson


On Tue, Feb 3, 2009 at 1:32 PM, Jonathon Anderson <janderso at alcf.anl.gov> wrote:
> Today, we will be upgrading clusterbank on both surveyor and intrepid.
> During the upgrade, qsub might fail to submit new jobs, due to the
> inability to check the status of your allocation during the upgrade.
>
> We estimate that the upgrade will take less than 1 hour. A separate
> notification will be sent out when the upgrade is complete.
>