Known issues
Intro
- Please report issues to support@computecanada.ca
- The status page at http://status.computecanada.ca/ is not updated automatically yet, so may lag in showing current status.
Scheduler errors
- The CC slurm configuration preferentially encourages whole-node jobs. Users should, if appropriate, request whole-node rather than per-core resources. See Job Scheduling - Whole Node Scheduling (Patrick Mann 20:15, 17 July 2017 (UTC))
- Cpu and Gpu backfill partitions have been created on both clusters. If a job is submitted with <24hr runtime, it will be automatically entered into the cluster-wide backfill partition. This partition has a low priority, but will allow increased utilization of the cluster by serial jobs. (Nathan Wielenga)
- SLURM epilog does not fully clean up processes from ended jobs, especially if the job did not exit normally. (Greg Newby) Fri Jul 14 19:32:48 UTC 2017)
- This has been greatly improved after the addition of the epilog.clean script, but there are still nodes occasionally marked down for epilog failure. (NW)
- By default, the job receives environment settings from the submitting shell. This can lead to irreproducible results if it's not what you expect. To force the job to run with a fresh-like-login environment, you can submit with --export=none or add #SBATCH --export=NONE to your job script.
Quota and filesystem problems
Quota errors on /project filesystem
Sometimes, users will see quota error on their project folders. This may happen when files are owned by a group other than the project group. You can change the group which owns files using the command
[name@server ~]$ chgrp -R <group> <folder>
To see what <group> should be, run the following command :
[name@server ~]$ stat -c %G $HOME/projects/*/
Only the owner of the files can run the chgrp command. To ask us to correct the group owner for many users, write to support@computecanada.ca
Nearline
- "Nearline" capabilities are not yet available (see https://docs.computecanada.ca/wiki/National_Data_Cyberinfrastructure for a brief description of the intended functionality)
- Update July 17: still not working. If you need your nearline RAC2017 quota then please ask CC support. (Patrick Mann 20:45, 17 July 2017 (UTC))
Missing symbolic links to project folders
- Upon login to the new clusters, symbolic links are supposed to be created in the user's account, as described in Project layout. Sometimes, it does not happen. If this is the case, please verify that your access to the cluster is enabled on this page https://ccdb.computecanada.ca/services/resources
Cedar only
- SLURM operations will occasionally time out with a message like "Socket timed out on send/recv operation" or "Unable to contact slurm controller (connect failure)". As a temporary workaround, attempt to resubmit your jobs/commands, they should go through in a few seconds. (Nathan Wielenga) 08:50, 18 July 2017 (MDT))
- Should be resolved after a VHD migration to a new backend for slurmctl. (NW)
Graham only
- /home is on an NFS appliance that does not support ACLs, so setfacl/getfacl doesn't work there.
- Workaround: use the /project or /scratch filesystems instead
- We're finding out whether this can be fixed through an update or reconfiguration.
- Compute nodes cannot access Internet
- Solution: Request exceptions to be made at support@computecanada.ca Describe what you need to access and why.
- Intel compiler does not work on compute nodes
- Solution/workaround: Compile your code on the login node.
Other issues
- modules don't work for shells other than bash(sh)
- Workaround (this appears to work but not tested extensively)
- source $LMOD_PKG/init/tcsh
- source $LMOD_PKG/init/zsh
- source $LMOD_PKG/init/ksh
- Bart will look at it in more detail
- Workaround (this appears to work but not tested extensively)