[temurin-compliance] AIX 7.2 machine restricted to 128 jobs per user

Summary

Our AIX 7.2 machine (jck-skytap-aix72-ppc64-3) is currently restricted to 128 processes per user as a hard limit. @fgurr has confirmed that this does not appear to be a limitation enforced in /etc/security/limits (although perhaps it's worth checking /etc/security/user too which we didn't do last week? They can't be accessed by non-root so I can't check myself)

$ ulimit -a | grep processes
processes(per user)  128
$ $ oslevel -s
7200-03-07-2114
$

The AIX 7.1 temurin compliance systems do not have the same restriction:

$ oslevel -s
7100-05-03-1846
$ ulimit -a | grep processes
processes(per user)  unlimited
$

Other AIX 7.2 systems I am using (albeit not at Skytap) are not showing the same restriction either:

jenkins@p8-aix1-adopt03:[/home/jenkins]oslevel -s
7200-04-02-2028
jenkins@p8-aix1-adopt03:[/home/jenkins]ulimit -a | grep processes
processes(per user)  unlimited
jenkins@p8-aix1-adopt03:[/home/jenkins]

so something on that box appears to have restricted it, but without root privileges I can't tell what it is (I assume root is not restricted)

Steps to reproduce

ulimit -a | grep processes

What is the current bug behavior?

Shows as 128 and cannot be extended, which is causing problems for some automated jobs for us unless we significantly lower the amount of parallelism that we are using

What is the expected correct behavior?

Max processes from ulimit is higher. Minimum of 256, ideally at least 512, but potentially unlimited to match the other systems.

Relevant logs and/or screenshots

All output is in the initial description

Priority

• Urgent
• High
• Medium
• Low

Severity

• Blocker
• Major
• Normal
• Low

Impact

(What is the impact of this issue? Is it blocking a release? Are there any time constraints?, for example: "We have a release tomorrow") in

We have a release cycle starting on Tuesday/Wednesday and while we have potential workarounds there is potential significant benefit to be derived from having the high process counts enabled in terms of execution time.

added Servers label

changed title from [temurin-compliace] AIX 7.2 machine restricted to 128 jobs per user to [temurin-compliance] AIX 7.2 machine restricted to 128 jobs per user

added IT-prioritymedium IT-severitynormal labels

AFAICT, /etc/security/user does not specify anything related to jobs/processes per user.

@jmazanek4ep can you check if you find something about why/how/where the number of processes per user are limited on the AIX 7.2 machine (skytap-aix72-ppc64-3)?

assigned to @jmazanek4ep

@saddisonxls all 3 machines have actually the same settings, the perceived difference is because between AIX 7.1 and AIX 7.2 the output of ulimit command changed.

In 7.1 it returns a value of system wide kernel process table limit - nproc, which by default is unlimited In 7.2 it returns a value of per user process limit - maxuproc, unless the nproc value is set in /etc/security/limits explicitly.

Value of maxuproc is currently set on all 3 VMs to 128 and can be displayed by lsattr command

That would make sense - thanks. Can we increase it from 128? ON the other AIX 7.2 system I have where it was returning unlimited it's got maxuproc set to 512, which is probably a better number for us. (I wonder if that needs a jenkins agent restart ... If so it's in use at the moment so give me a shout before restarting)

I have increased it to 512 on 7.2 machine. Restart shouldn't be necessary

bash-5.1# lsattr -E -l sys0 -a maxuproc
maxuproc 512 Maximum number of PROCESSES allowed per user True

LGTM - it doesn't take effect on existing processes but anything forked from them get the new value (I have an existing shell - it still shows 128, but if I start another process it get the 512 ulimit so that should be good - thanks!)

closed