Wolfram - Matematica
( update by David )
https://cleo.mit.edu/issues/6226
* David got contacted by
* we can use a central licence, users do not need the licence
* Wolfram udpated on all the login nodes (not only 00) (david)
* jupiter configuration need to be updated with all the login nodes (Matt)
* future: check the slurm nodes i.e. if users need to send longer jobs on slurm
* stale: some version is here not used /work/submit/sysadmin/config/submit.mit.edu/
----------------------
login nodes configuration:
( update by David )
* submit00 was often stuck
* there was a throttling in place that included root and users
--> root treated separately
--> removed the throttling, jobs above some quota get killed.
--> set memory limit 12 GB per user (before there was also a total memory limit)
* TODO: document this on a cleo ticket
----------------------
management network set-up:
( update by David )
5 out of 8 login nodes have it
to update https://github.com/mit-submit/submit-admins-guide/blob/main/source/management_interface.rst
( in the future have a uniform password )
----------------------
quotas:
( update by David )
new command 'quota -s' so that user can directly query
[mariadlf@submit82 ~]$ quota -s
Disk quotas for user mariadlf (uid 146312):
Filesystem space quota limit grace files quota limit grace
submit-filer-02-clus1.mit.edu:/submit_filer_2
195G 199G 200G 983k 10000k 11000k
submit-filer-02-clus1.mit.edu:/submit_filer
20253M 20480M 21504M 410k* 400k 500k
----------------------
LHCb & xRootD:
( update by David )
https://cleo.mit.edu/issues/6224
LHCB people want to read from CEPH from xrootD
--> created a group
--> can use the address xrootD submit50 ... 59
==> in the future get the redirector
----------------------
Matt:
1) simplify the script to query users, discussed the old users
2) alarms: will work on aggregate metrics i.e. how many nodes are out of slurm ...
--> later reorganize some alarms in different channels