[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Condor-users] Avoiding Cross-Flocking
- Date: Thu, 30 Apr 2009 16:21:55 +0200
- From: Colak Birol <birol.colak@xxxxxxxxx>
- Subject: Re: [Condor-users] Avoiding Cross-Flocking
instead of changing the FLOCK_TO and FLOCK_FROM parameters on the fly
(that seems not to be working), I finally have a
I add all computing nodes the machine classad flocking_enabled and give
the jobs the requirement
(flocking_enabled == TRUE || COLLECTOR_HOST_STRING == "hpc1mn").
Where hpc1mn is the host from where the jobs are being submitted. With a
cronjob I check if the local central manager flocks. In that case I
reconfigure the local pool setting flocking_enabled = FALSE, so that no
remote jobs can flock to the local pool.
I have configured following in the central managers local configuration
You could run a cron job on each of your pools that changes FLOCK_TO and
FLOCK_FROM depending on what's going on in your system. If query your
schedds for one current pool, look at the values of TotalFlockedJobs,
you can set FLOCK_FROM for machines in that pool to be empty if any
schedd has TotalFlockedJobs > 0.
I'm not sure if FLOCK_TO or FLOCK_FROM can contain ClassAd expression
logic. If they can you might be able to do something fancier using
ClassAd expressions and some logic right in the setting. See:
I'm not a user of flock technology -- maybe someone who uses it has a
more elegant way to do it. But that's the first brute force method that
comes to mind.
ENABLE_RUNTIME_CONFIG = True
ENABLE_PERSISTENT_CONFIG = True
PERSISTENT_CONFIG_DIR = /opt/condor/local/persistent_config
SETTABLE_ATTRS_CONFIG = *
SETTABLE_ATTRS_OWNER = flock_to, flock_from, flocking_partner
And I configured in the global configuration file:
FLOCK_FROM = $(flocking_partner)
FLOCK_TO = $(flocking_partner)
condor_config_val -master -set "flocking_partner=xxxx"
condor_config_val -master -rset "flocking_partner=xxxx"
condor_reconfig -master -full
1. It seems that I can not change the FLOCK_FROM and FLOCK_TO during
2. I have always TotalFlockedJobs = 0 (also when there is flocking, I
checked with command: condor_status -schedd -l | grep TotalFlockedJobs).
Condor-users mailing list
To unsubscribe, send a message to condor-users-request@xxxxxxxxxxx with a
You can also unsubscribe by visiting
The archives can be found at: