Holiday file is out of date

Users/Site Administrators
1

I’m seeing a lot of this in the logs:

11/11/2019 15:23:55;0004;pbs_sched;Fil;holidays;The holiday file is out of date; please update it.

We have Year 0 in our conf file. According to this comment from 2016, we can have Year 0 but there will be an error message in the logs.

Is that the message, and is there any way to not have a holiday file? I note that there are other solutions here but nothing that both removes the need to have a holiday file and removes the log message?

2

holidays file is required for the prime/non-prime time

Please comment the entire holidays file and kill -HUP or restart pbs services.
This will make sure the message is not seen in the sched logs . This will make everything is prime time

1 Like
3

Thanks! Am investigating now the reasons for having prime time.

4

We checked in some enhancements to the holidays file feature some time back: https://github.com/PBSPro/pbspro/pull/979

Setting year to 0 should also not cause it to print the “holidays file is out of date” message anymore.

But maybe you’ll find use of the prime-time feature, so do check that out.

1 Like
5

That’s great - which version of PBSPro does that ship in? By my reckoning (a quick visual analysis of release dates) it was merged after v19.1.1 was released. Thanks

6

This commit contains the above PR :

* commit 9e03b31cb3355304969b6afccf1e78d709231757
| Author: Ravi Agrawal <ragrawal@altair.com>
| Date:   Mon Feb 4 16:36:19 2019 +0530
| 
|     holidays file: 24hrs prime-time for missing/inconsistent days
|     
|     Conflicts:
|       src/scheduler/pbs_holidays

If you use previous version of PBS and building from source, You have some choices.

In our cluster, I use PBSPro v18.1.4 and building from source code (cloned by github),
(and v19.x did not work), so I did:

$ git fetch --all
$ git cherry-pick 9e03b31cb3355304969b6afccf1e78d709231757

Best,

7

Hey,

Sorry for the late reply. Sounds like you were able to cherry-pick the commit, hopefully it works out ok.
Can you tell me what issues you came across when building v19.x from source?

Thanks,
Ravi

8

Hi @agrawalravi90 !

Build is successfully.
I submit 1 job, then the node status becomes ‘job-busy’, even if the node have more free cpus.
This problem also happens on 18.1.x, but
I know that I can correct this problem by restarting pbs_server (sudo service pbs restart).
But In 19.x, I couldn’t correct it even if I restarted server.

And now I try to investigate the minimal configurations which reproduces problem (and then I’d like to post forum if I find reproduce conditions).

By gitting bisect, the following commit raises this ?:

* commit e270920b3d6cb8912f0bb420ee2821bea0e201e8 (HEAD, refs/bisect/bad)
| Author: Suresh Thelkar <suresh.thelkar...
| Date:   Thu Aug 16 16:40:33 2018 +0530
|
|     Database refactoring, creation and using API for the same
9

Thanks for explaining the issue @Takahiro. Just to be sure, the node’s ‘sharing’ attribute is set to ‘default_shared’ right? That should be the value that it gets out of the box.

@suresht might know more about the commit that you found. Although, I suspect that the commit just made it so that some information that was previously not stored to the database was now stored, that’s why you still see the issue after a server restart.