[LU-12330] Support per-session JOBID without using the environment

• Type: Improvement
• Resolution: Fixed
• Priority: Minor
• Fix Version/s: Lustre 2.13.0
• Assignee: Neil Brown
• Reporter: Neil Brown

Description

Reading from the environment of a process is something only that process should do. The current jobid code does this in the kernel and there is good reason to expect resistance to this when we resubmit the code for upstream inclusion.

So we need a different way to provide a job-id which is common to all tasks in a job across a cluster, but not necessarily all tasks on any given node.

A proposed way to do this is to leverage the "session-id" concept (see setsid(1)).

If we can associate a jobid with a session-id, then every session servicing a given job can be given the same jobid. The kernel can easily get the session-id for a process, and can then use an internal mapping to map this to the correct jobid.

 

Issue Links

• is duplicated by: LU-4186 Change lustre_get_jobid to read from proc file and cache it in lu_env
• is related to: LU-12511 Prepare lustre for adoption into the linux kernel
• is related to: LU-17512 add conditional operator for 'jobid_name'
• is related to: LU-10698 Specify complex JobIDs for Lustre