gridengine.info : Category maillist-bits, everything about maillist-bits

Fedora 10 will ship with SGE 6.2

dag@sonsorol.org (chris) — Mon, 17 Nov 2008 18:36:30 -0500

I'm late in catching up with Grid Engine mailing list traffic but this one from Orion Poplawski caught my eye:

F-10 will ship with 6.2-3. I'll be pushing a 6.3-4 (or later) 0-day update as well: * Tue Nov 11 2008 - Orion Poplawski - 6.2-4 - Add note to README about localhost line in /etc/hosts - Cleanup setting.sh some, no more MAN stuff - Add conditional build support for EL - Use system db_* utils in bdb_checkpoint script

I've got the src.rpm here: http://www.cora.nwra.com/~orion/fedora/gridengine-6.2-4.fc11.src.rpm This should build on EL-5, F-9, and F-8 with sun java 1.6.0 installed. These rpms are geared for minimal NFS type installs. install_* scripts should work, though install_execd should not be needed for standard "default" installs. Bugs to https://bugzilla.redhat.com/.

Fedora 10 will ship with SGE 6.2

dag@sonsorol.org (chris) — Mon, 17 Nov 2008 18:36:30 -0500

I'm late in catching up with Grid Engine mailing list traffic but this one from Orion Poplawski caught my eye:

Reuti: Tight integration with Intel MPI 3.1 or MPICH2

dag@sonsorol.org (chris) — Mon, 17 Nov 2008 17:14:46 -0500

Via this thread ...

Reuti has updated his methods and information for achieving tight integration in MPICH2 environments. An updated set of files for mpd integration for MPICH(2) is now at http://gridengine.sunsource.net/howto/mpich2-integration/mpich2-60.tgz

The thread discusses Intel MPI 3.1 with the suggestion that the above methods for MPICH2 may work with the Intel product. The basic issue is that the standard "mpdboot" method has always been difficult to achieve tight integration with Grid Engine environments.

Fixing a berkeley db spool database

dag@sonsorol.org (chris) — Tue, 11 Nov 2008 12:43:00 -0500

Per this thread on the users list, a recepie for rebuilding and re-verifying a Berkeley based binary SGE spool:


service sgemaster stop # on failover server
service sgemaster stop # on master server

cd $SGE_ROOT/default/spool
cp -a spooldb spooldb.bak

cd spooldb
$SGE_ROOT/utilbin/lx24-amd64/db_verify sge
$SGE_ROOT/utilbin/lx24-amd64/db_recover
$SGE_ROOT/utilbin/lx24-amd64/db_dump -f sge.out sge
mv sge sge.old
$SGE_ROOT/utilbin/lx24-amd64/db_load -f sge.out sge
$SGE_ROOT/utilbin/lx24-amd64/db_verify sge


service sgemaster start # on master server
service sgemaster start # on failover server

Fixing a berkeley db spool database

dag@sonsorol.org (chris) — Tue, 11 Nov 2008 12:43:00 -0500

Per this thread on the users list, a recepie for rebuilding and re-verifying a Berkeley based binary SGE spool:


service sgemaster stop # on failover server
service sgemaster stop # on master server

cd $SGE_ROOT/default/spool
cp -a spooldb spooldb.bak

cd spooldb
$SGE_ROOT/utilbin/lx24-amd64/db_verify sge
$SGE_ROOT/utilbin/lx24-amd64/db_recover
$SGE_ROOT/utilbin/lx24-amd64/db_dump -f sge.out sge
mv sge sge.old
$SGE_ROOT/utilbin/lx24-amd64/db_load -f sge.out sge
$SGE_ROOT/utilbin/lx24-amd64/db_verify sge


service sgemaster start # on master server
service sgemaster start # on failover server

Grid Engine, workflows & virtualization

dag@sonsorol.org (chris) — Fri, 07 Nov 2008 13:04:10 -0500

Another discussion happening recently on the SGE user list concerns how best to handle virtualization. That thread can be browsed here.

In a followup, Andreas is soliciting feedback from the wider community on how you want to see this area handled in future revisions of Grid Engine. Time to speak up if you have an opinion!

Read Andreas's request for feedback here.

Grid Engine & power saving

dag@sonsorol.org (chris) — Fri, 07 Nov 2008 12:02:17 -0500

I'd guess that most people don't follow the SGE developer list all that closely. Sometimes the developer discussions cross over into areas that all users may be interested in.

There has been an interesting discussion on various ways to give SGE the ability to either directly trigger or otherwise interact with various systems that either switch nodes down into lower power states or even completely power them down/up as needed (Project Hedeby / SDM, etc.)

Automatic methods for powering up and down portions of clusters based on workload have been used for years now but the topic seems to be getting more interest and more backing. A few years ago I saw a neat solution that some people at Cornell Medical College had done -- they used PBS/Torque and had various IPMI scripts that powered nodes on or off depending on the size of the pending job list.

The developer thread (via MarkMail) is here. The CollabNet "Forum View" is here.

Help make grid engine better

dag@sonsorol.org (chris) — Tue, 23 Sep 2008 11:33:29 -0400

The Grid Engine development team can't develop the product in a vacuum -- feedback, suggestions and input from real-world production users of Grid Engine is always critical.

If you are a serious Grid Engine user and care about the future direction of software development efforts, take the time to read this proposal for a new "Job Submission Verifier" enhancement. If the subject is of interest to how you use SGE, the developers would welcome comments, suggestions and feedback before Friday.

"Job Submission Verifier Specifications:"
http://gridengine.sunsource.net/servlets/ReadMsg?list=users&msgNo=25999

Follow the email thread via MarkMail if you are interested in what others are saying about this.

Help make grid engine better

dag@sonsorol.org (chris) — Tue, 23 Sep 2008 11:33:29 -0400

The Grid Engine development team can't develop the product in a vacuum -- feedback, suggestions and input from real-world production users of Grid Engine is always critical.

"Job Submission Verifier Specifications:"
http://gridengine.sunsource.net/servlets/ReadMsg?list=users&msgNo=25999

Follow the email thread via MarkMail if you are interested in what others are saying about this.

Fixing SGE email issues on Apple OS X

dag@sonsorol.org (chris) — Tue, 23 Sep 2008 10:57:31 -0400

Are you in the following situation?

/usr/bin/mail works perfectly from the command line
/usr/bin/mail configured as the SGE mailer produces no email
substituting a wrapper with extra logging also produces no logs or email

The only clue is in the spool logs:

09/10/2008 16:22:07|execd|xxx-fs01|E|mailer had timeout - killing
09/10/2008 16:22:07|execd|xxx-fs01|E|mailer exited with exit status= 1
09/10/2008 16:22:19|execd|xxx-fs01|E|mailer had timeout - killing
09/10/2008 16:22:19|execd|xxx-fs01|E|mailer exited with exit status= 1

Thanks to Valerio Luccio we have a workaround. The issue is apparently a conflict between one of the SGE supplied libraries that interferes with the mail MTA on OS X when SGE tries to invoke it. A trivial wrapper script that overrides the DYLD_LIBRARY_PATH environment variable is the fix:

#!/bin/sh
export DYLD_LIBRARY_PATH=/usr/lib
/usr/bin/mail -s "$2" $3

This solved a problem that had been bothering me for days, thanks Valerio - I owe you a beer if we ever end up at the same meeting or conference!

Fixing SGE email issues on Apple OS X

dag@sonsorol.org (chris) — Tue, 23 Sep 2008 10:57:31 -0400

Are you in the following situation?

/usr/bin/mail works perfectly from the command line
/usr/bin/mail configured as the SGE mailer produces no email
substituting a wrapper with extra logging also produces no logs or email

The only clue is in the spool logs:

09/10/2008 16:22:07|execd|xxx-fs01|E|mailer had timeout - killing
09/10/2008 16:22:07|execd|xxx-fs01|E|mailer exited with exit status= 1
09/10/2008 16:22:19|execd|xxx-fs01|E|mailer had timeout - killing
09/10/2008 16:22:19|execd|xxx-fs01|E|mailer exited with exit status= 1

#!/bin/sh
export DYLD_LIBRARY_PATH=/usr/lib
/usr/bin/mail -s "$2" $3

This solved a problem that had been bothering me for days, thanks Valerio - I owe you a beer if we ever end up at the same meeting or conference!

MarkMail: Mine the grid engine maillist archives

dag@sonsorol.org (chris) — Wed, 17 Sep 2008 07:48:31 -0400

MarkMail has just imorted all of the Grid Engine mailing lists from http://gridengine.sunsource.net into their archive, search, index and database system. Initial results are pretty impressive based on a few minutes of searching and experimentation -- seems like a great way to search the mailing lists for answers and info.

Click on the image above and you'll be take to a search on the term 'rqs'. Leave a comment with your impressions if you are so inclined.

MarkMail: Mine the grid engine maillist archives

dag@sonsorol.org (chris) — Wed, 17 Sep 2008 07:48:31 -0400

Click on the image above and you'll be take to a search on the term 'rqs'. Leave a comment with your impressions if you are so inclined.

Bugfix madness

dag@sonsorol.org (chris) — Fri, 12 Sep 2008 07:51:46 -0400

Wow, this is what my inbox looked like this morning -- a massive influx of resolved SGE issues via the SGE issues mailing list. Go Jana!

Feedback needed: Obsolete options and parameters considered for removal

dag@sonsorol.org (chris) — Tue, 24 Jun 2008 08:22:41 -0400

Grid Engine developers posted a list today of SGE configuration parameters and client arguments that are being considered for removal from the product because they are either obsolete or they duplicate settings found elsewhere.

The developers are seeking feedback and comments on their plans - if you have any please drop a line to the users@gridengine.sunsource.net mailing list. The current roadmap calls for these methods to be marked as 'deprecated' in the SGE 6.2 release with total removal planned for a future post-6.2 release.

The message can be found here:
http://gridengine.sunsource.net/servlets/ReadMsg?list=users&msgNo=25045

The full list of items being considered for removal can also be found after the jump ...

The parameters planned to obsolete are:

host_conf(5)
- processors
   obsolete, same as num_proc from the complex list

sched_conf(5)
- algorithm
   just default is allowed, no additional algorithms are planed
- params JC_FILTER
   huge performance impact plus may lead to wrong scheduling decisions

sge_conf(5)
- reprioritize
   redundant because hard bound to reprioritize_interval in sched_conf(5)
- shell_start_mode
   obsolete, value from queue_conf(5) is used
- set_token_cmd
   no known AFS support
- pag_cmd
   no known AFS support
- token_extend_time
   no known AFS support
- qmaster_params DISABLE_AUTO_RESCHEDULING
   equivalent to default reschedule_unknown=0:0:0
- qmaster_params merge ACCT_RESERVED_USAGE and SHARETREE_RESERVED_USAGE
   We can't imaging a use case to have these values separated
- finished_jobs
   qstat -j does not work with successful finished jobs. Code seems
   to work only with jobs going into error state.

user(5)
- delete_time
   change to internal, not changeable/visible field
   Implicit set by auto_user_delete_time

qconf(1)
- sep option
   obsolete, same as num_proc
- ks option
   obsolete, same as -kt scheduler

qmod(1)
- c option
   depreciated, use -cj or -cq
- r option
   depreciated, use -rj or -rq
- s option
   depreciated, use -sj or -sq
- us option
   depreciated, use -usj or -usq