Hi David, as long as you don't do any seeks in the output streams, maybe pipes could work, i.e., create locally in each job/worker named pipes to write the streams to - and then read and pipe the named pipe through zcat compressed on to the shared file system. A problem will probably, that the streams might already produce output before a wrapper could set up pipes - maybe a preCmd could work?? (have not tried it tbh) Cheers, Thomas On 18/04/2022 22.52, David Schultz wrote: > Hi Stuart, > > I do compressed stdout/err as part of my jobs on the execute side, > partly to decrease network traffic through the schedd. To avoid > malformed or incomplete log files, I use a wrapper script around my > actual job payload which redirects the payload stdout/err to different > files, compresses them at the end of the job, and transfers those files > back explicitly. The only things in my HTCondor stdout/err are from my > wrapper script for setting up the job environment. > > I've tried several more clever things over the years, but have had > issues with them. This is very simple and robust. > > Best, > David > > > On Mon, Apr 18, 2022 at 1:57 PM Greg Thain <gthain@xxxxxxxxxxx > <mailto:gthain@xxxxxxxxxxx>> wrote: > > On 4/18/22 05:03, Stuart Walker wrote: > > Hello, > > > > I have a slight problem that I wondered if someone could help me with. > > > > My error, log and output files are set like this in my job > submission script: > > > > output       = > /afs/desy.de/user/s/stwalker/luxe-bb-bdsim-model/bdsim-models/beam-halo-model-2022-04-12/t20-halo-from-distr-file/output/out/bdsim-$(Cluster)_$(Process).out > <http://desy.de/user/s/stwalker/luxe-bb-bdsim-model/bdsim-models/beam-halo-model-2022-04-12/t20-halo-from-distr-file/output/out/bdsim-$(Cluster)_$(Process).out> > > error       Â= > /afs/desy.de/user/s/stwalker/luxe-bb-bdsim-model/bdsim-models/beam-halo-model-2022-04-12/t20-halo-from-distr-file/output/error/bdsim-$(Cluster)_$(Process).error > <http://desy.de/user/s/stwalker/luxe-bb-bdsim-model/bdsim-models/beam-halo-model-2022-04-12/t20-halo-from-distr-file/output/error/bdsim-$(Cluster)_$(Process).error> > > log        Â= > /afs/desy.de/user/s/stwalker/luxe-bb-bdsim-model/bdsim-models/beam-halo-model-2022-04-12/t20-halo-from-distr-file/output/log/bdsim-$(Cluster)_$(Process).log > <http://desy.de/user/s/stwalker/luxe-bb-bdsim-model/bdsim-models/beam-halo-model-2022-04-12/t20-halo-from-distr-file/output/log/bdsim-$(Cluster)_$(Process).log> > > > Hi Stuart: > > Looking at the log file first, there shouldn't be too many bytes > written > to it per job, so I hope that's not a problem for your quota. HTCondor > does support safely sharing one big log file for many jobs -- perhaps > that would help with minimizing the number of files or inodes. You > could set > > log = > /afs/desy.de/user/s/stwalker/luxe-bb-bdsim-model/bdsim-models/beam-halo-model-2022-04-12/t20-halo-from-distr-file/output/log/bdsim-some-workflow-name.log > <http://desy.de/user/s/stwalker/luxe-bb-bdsim-model/bdsim-models/beam-halo-model-2022-04-12/t20-halo-from-distr-file/output/log/bdsim-some-workflow-name.log> > > Then all the jobs in some workflow that share that log file will safely > append to it. > > > Now, for the output and error files, I don't think we have a good > way in > HTCondor proper to compress them. If these jobs are part of a DAG, > then > a DAG postscript (which runs on the submit machine) could compress the > files. However, I suspect it is better overall to explicitly compress > them as part of the job on the execute side. > > > -greg > > _______________________________________________ > HTCondor-users mailing list > To unsubscribe, send a message to htcondor-users-request@xxxxxxxxxxx > <mailto:htcondor-users-request@xxxxxxxxxxx> with a > subject: Unsubscribe > You can also unsubscribe by visiting > https://lists.cs.wisc.edu/mailman/listinfo/htcondor-users > <https://lists.cs.wisc.edu/mailman/listinfo/htcondor-users> > > The archives can be found at: > https://lists.cs.wisc.edu/archive/htcondor-users/ > <https://lists.cs.wisc.edu/archive/htcondor-users/> > > > _______________________________________________ > HTCondor-users mailing list > To unsubscribe, send a message to htcondor-users-request@xxxxxxxxxxx with a > subject: Unsubscribe > You can also unsubscribe by visiting > https://lists.cs.wisc.edu/mailman/listinfo/htcondor-users > > The archives can be found at: > https://lists.cs.wisc.edu/archive/htcondor-users/
Attachment:
smime.p7s
Description: S/MIME Cryptographic Signature