Migrate the control node

[wakaka]

I plan to migrate the control node. How can I retain the original usage records and other information?

[adarsh]

1. Take a backup of the configuration and files

/etc/pbs.conf
$PBS_HOME/sched_priv/sched_config
qmgr -c "print server"  > printserver.txt
qmgr -c "print sched" > printsched.txt
qmgr -c "print nodes @default" > printnodes.txt
qmgr -c "print queue @default" > printqueue.txt
qmgr -c "print hook @default" > printhook.txt
Note down the last jobid
backup any custom scripts or hooks

2. stop the pbs services on the control node
   backup $PBS_HOME , $PBS_EXEC , /etc/pbs.conf , /etc/init.d/pbs , /etc/profile.d/pbs.*

3. Bring up a new VM , note : the control node hostname and compute node hostname should be same in the /etc/hosts of this new VM.

install  PBS Pro 
start pbs pro services # do not submit any jobs 
migrate the configuration from the backup (qmgr < printserver.txt and so on )
copy the accounting logs from backup $PBS_HOME/server_priv  to $PBS_HOME/server_priv
edit the postgres and update the next jobid to be   last jobid+1
submit a test job

Please go through this guide : https://2021.help.altair.com/2021.1/PBSProfessional/PBSInstallGuide2021.1.pdf

[wakaka]

How can I update the next jobid to be last jobid+1?

[adarsh]

Please check the protocol here

OpenPBS database corrupted

Please find my protocol below: [root@pbsserver ~]# qstat --version pbs_version = 23.06.06 [root@pbsserver ~]# pbs_ds_password (Say you have the password set: openpbs ) Enter the password: Re-enter the password: ---> Updated user password ---> Success [root@pbsserver ]# ps -ef | grep pbs_ root 126459 1 0 08:43 ? 00:00:00 /opt/pbs/sbin/pbs_comm root 126474 1 0 08:43 ? 00:00:00 /opt/pbs/sbin/pbs_sched root 126583 1 0 08:43 ? 00:00:00 /op…

[wakaka]

Some jobs are in the queue and are being executed on the compute nodes (state R), while many other jobs are queued (state Q). How can I restore the system to its original state after the migration?
I encountered another problem: my compute node is completely out of disk space. PBS submits the job to this node, but it remains in the R state. I checked the logs and it seems to be repeatedly trying. Can’t PBS just terminate the job directly?

[adarsh]

@wakaka

Some jobs are in the queue and are being executed on the compute nodes (state R), while many other jobs are queued (state Q). How can I restore the system to its original state after the migration?

• You cannot save the running jobs during migration. You have to set (qmgr: set server scheduling = false) and allow the running jobs to complete.
• Also, it is better to ignore the queued and running jobs (and drain them) and do the migration
• stop PBS services on the Server host and take a postgres dump
• import the dump into the new server
• copy the accounting logs
• start the PBS services

Note: The migration discussed above was reconfiguring the new server node with respect to configuration and accounting logs and the next job id.