Removing a PBS server and adding it as a MoM to an existing other PBS server

#1

Hi PBS-community,

we have two very small cluster systems each running a different PBS version:

  1. Cluster: Login node + 2 MoMs (PBS server runs on the login node)
  2. Cluster: 1 MoM (PBS server and MoM on the same node)

We want to add the second cluster as a MoM to the PBS server which is running on the login and add a new queue so the users can use both systems from a single login node.

I have several questions regarding this change.

  1. The login node runs Red Hat Enterprise Linux 7.4 and the second cluster uses Ubuntu 18.04.1 LTS, can the difference in operating systems cause any major obstacle?

  2. Is there any way to disable the PBS server on the second cluster without removing the server and configuration files, because if the configuration does not work correctly we want to continue to use the current configuration with two PBS servers?

  3. The login node runs a PBS server 14.2.1, what is the easiest way to add the second cluster as a MoM?

Thank you for your help

#2

@yali Which version of PBS mom is running on the 2nd cluster(the one you wish to add to the 14.2.1 server) ?

#3

Hi visheshh,

thank you for your answer!
The second cluster runs the Open Source version 19.0.0 of PBS.

#4

PBS Pro major version should be same across the entire PBS complex. We can mix and match Linux flavour of operating systems by running the same version of PBS Pro , however recommendation is to use the same flavour & version of Linux operating system across the entire complex

  1. Stop the PBS Services on the second cluster
    2 . edit /etc/pbs.conf and update it to below values for these variables and start the services
    PBS_START_SERVER=0
    PBS_START_SCHED=0
    PBS_START_COMM=0
    PBS_START_MOM=1

Note: Atleast the major version of PBS Pro should be same, better to upgrade to the latest.

  1. Make sure to update the PBS_SERVER in pbs.conf and mom_priv/config’s $clienthost , so that they point to the login node PBS Server
  2. Start the PBS Services
  3. If there are any issues, then stop the PBS Services , revert back these values to what they were earlier and start the pbs services.
#5

Hi adarsh,

thank you for the answer!
By major version you mean, we should update from 14.2.1 PBS Pro on the login node to the Open Source version 19.0.0?
Can the PBS Pro version easily be replaced by the Open Source version?

#6

Update the 14.2.1 PBS Pro on the login node to 19.2.1 which will be released by the end of next month. Also, do not mix and match the Pro and OSS version on your setup.

You have to uninstall Pro (take a pbs_diag backup before uninstall) and install OSS and re-configure it with the backup configuration files. If you are in production and on PBS Pro support, then continue with that and upgrade once the 19.2.1 is released.

#7

Thank you for your answer!
Is it recommended to wait for the 19.2.1 release and not use the current 19.0.0 release?

#8

I do not recommend 14.2.1 server with 19.0.0 mom . As Adarsh mentioned please use 19.2 server.

1 Like