Skip to content

Add missing drive name to taped process names

At the moment, the process name of the maintenance process is set to cta-tpd-maint. If multiple taped services are running on the same machine, it means that the process name cannot be used to uniquely identify which taped service it belongs to. A similar problem appears for the master process, which is cta-tpd-master.

This is particularly problematic for telemetry, where the process name will (need to) be used to uniquely identify a process on a given host.

Additionally, it is also problematic for operations. Executing ps -u cta allows operators to see the processes and potentially kill one of them if needed. However, if multiple taped processes are running, there is no distinction between the maint/master processes.

For reference, here is what is shown on a tapeserver with 2 drives actively transferring:

[root@tpsrv640 ~]# ps -eo pid,user,comm | grep cta
 726964 cta      cta-rmcd
3262030 cta      cta-tpd-master
3262033 cta      cta-tpd-maint
3262257 cta      cta-tpd-master
3262259 cta      cta-tpd-maint
3702349 cta      cta-tpd-F10B1S2
3708150 cta      cta-tpd-F10B1S1

Simply appending the drive name is not possible, as there is a 16-byte limit to the process name.

We should make sure this is consistent. Doing so is cleaner and allows us to better identify the processes.


Accepted Proposal

To close down the discussion, we are going for proposal 3:

F10B1S2-parent
F10B1S2-maint
F10B1S2-drive

There will be a check on the drive name part to ensure it remains below 13 characters. This way, we still keep a single letter from the postfix allowing the process to be identified. The check will produce a warning (not an error), as we don't want to break existing deployments for something that does not concretely affect any functionality.

Edited by Niels Alexander Buegel