[slurm-users] Slurm Upgrade

Mon Nov 2 15:31:49 UTC 2020

We don't follow the recommended procedure here but rather build RPMs and 
upgrade using those.  We haven't and any issues.  Here is our procedure:

1. Build rpms from source using a version of the slurm.spec file that we 
maintain. It's the version SchedMD provides but modified with some 
specific stuff for our env and to disable automatic restarts on upgrade 
which can cause problems especially for upgrading the slurm database.

2. We test the upgrade on our test cluster using the following sequence.

a. Pause all jobs and stop all scheduling.

b. Stop slurmctld and slurmdbd.

c. Backup spool and the database.

d. Upgrade slurm rpms (note that you need to make sure that the upgrade 
will not automatically restart the dbd or the ctld else you may end up 
in a world of hurt)

e. Run slurmdbd -Dvvvvv to do the database upgrade. Depending on the 
upgrade this can take a while because of database schema changes.

f. Restart slurmdbd using the service

g. Upgrading slurm rpms across the cluster using salt.

h. Global restart of slurmd and slurmctld.

3. If that all looks good we rinse and repeat on our production cluster.

The rpms have worked fine for us.  The main hitch is the automatic 
restart on upgrade, which I do not recommend.  You should neuter that 
portion of the provided spec file, especially for the slurmdbd upgrades.

We generally prefer the RPM method as it is the normal method for 
interaction with the OS and works well with Puppet.

-Paul Edmon-

On 11/2/2020 10:13 AM, Jason Simms wrote:
> Hello all,
>
> I am going to reveal the degree of my inexperience here, but am I 
> perhaps the only one who thinks that Slurm's upgrade procedure is too 
> complex? Or, at least maybe not explained in enough detail?
>
> I'm running a CentOS 8 cluster, and to me, I should be able simply to 
> update the Slurm package and any of its dependencies, and that's it. 
> When I looked at the notes from the recent Slurm Users' Group meeting, 
> however, I see that while that mode is technically supported, it is 
> not recommended, and instead one should always rebuild from source. 
> Really?
>
> So, ok, regardless whether that's the case, the upgrade notes linked 
> to in the prior post don't, in my opinion, go into enough detail. It 
> tells you broadly what to do, but not necessarily how to do it. I'd 
> welcome example commands for each step (understanding that changes 
> might be needed to account for local configurations). There are no 
> examples in that section, for example, addressing recompiling from source.
>
> Now, I suspect a chorus of "if you don't understand it well enough, 
> you shouldn't be managing it." OK. Perhaps that's fair enough. But I 
> came into this role via a non-traditional route and am constantly 
> trying to improve my admin skills, and I may not have the complete 
> mastery of all aspects quite yet. But I would also say that 
> documentation should be clear and complete, and not written solely for 
> experts. To be honest, I've had to go to lots of documentation 
> external to SchedMD to see good examples of actually working with 
> Slurm, or even ask the helpful people on this group. And I firmly 
> believe that if there is a packaged version of your software - as 
> there is for Slurm - that should be the default, fully-working way to 
> upgrade.
>
> Warmest regards,
> Jason
>
> On Mon, Nov 2, 2020 at 9:28 AM Paul Edmon <pedmon at cfa.harvard.edu 
> <mailto:pedmon at cfa.harvard.edu>> wrote:
>
>     In general  I would follow this:
>
>     https://slurm.schedmd.com/quickstart_admin.html#upgrade
>     <https://slurm.schedmd.com/quickstart_admin.html#upgrade>
>
>     Namely:
>
>     Almost every new major release of Slurm (e.g. 19.05.x to 20.02.x)
>     involves changes to the state files with new data structures, new
>     options, etc. Slurm permits upgrades to a new major release from
>     the past two major releases, which happen every nine months (e.g.
>     18.08.x or 19.05.x to 20.02.x) without loss of jobs or other state
>     information. State information from older versions will not be
>     recognized and will be discarded, resulting in loss of all running
>     and pending jobs. State files are *not* recognized when
>     downgrading (e.g. from 19.05.x to 18.08.x) and will be discarded,
>     resulting in loss of all running and pending jobs. For this
>     reason, creating backup copies of state files (as described below)
>     can be of value. Therefore when upgrading Slurm (more precisely,
>     the slurmctld daemon), saving the /StateSaveLocation/ (as defined
>     in /slurm.conf/) directory contents with all state information is
>     recommended. If you need to downgrade, restoring that directory's
>     contents will let you recover the jobs. Jobs submitted under the
>     new version will not be in those state files, but it can let you
>     recover most jobs. An exception to this is that jobs may be lost
>     when installing new pre-release versions (e.g. 20.02.0-pre1 to
>     20.02.0-pre2). Developers will try to note these cases in the NEWS
>     file. Contents of major releases are also described in the
>     RELEASE_NOTES file.
>
>     So I wouldn't go directly to 20.x, instead I would go from 17.x to
>     19.x and then to 20.x
>
>     -Paul Edmon-
>
>     On 11/2/2020 8:55 AM, Fulcomer, Samuel wrote:
>>     We're doing something similar. We're continuing to run production
>>     on 17.x and have set up a new server/cluster  running 20.x for
>>     testing and MPI app rebuilds.
>>
>>     Our plan had been to add recently purchased nodes to the new
>>     cluster, and at some point turn off submission on the old cluster
>>     and switch everyone to submission on the new cluster (new
>>     login/submission hosts). That way previously submitted MPI apps
>>     would continue to run properly. As the old cluster partitions
>>     started to clear out we'd mark ranges of nodes to drain and move
>>     them to the new cluster.
>>
>>     We've since decided to wait until January, when we've scheduled
>>     some downtime. The process will remain the same wrt moving nodes
>>     from the old cluster to the new, _except_ that everything will be
>>     drained, so we can move big blocks of nodes and avoid slurm.conf
>>     Partition line ugliness.
>>
>>     We're starting with a fresh database to get rid of the bug
>>     induced corruption that prevents GPUs from being fenced with cgroups.
>>
>>     regards,
>>     s
>>
>>     On Mon, Nov 2, 2020 at 8:28 AM navin srivastava
>>     <navin.altair at gmail.com <mailto:navin.altair at gmail.com>> wrote:
>>
>>         Dear All,
>>
>>         Currently we are running slurm version 17.11.x and wanted to
>>         move to 20.x.
>>
>>         We are building the New server with Slurm 20.2 version and
>>         planning to upgrade the client nodes from 17.x to 20.x.
>>
>>         wanted to check if we can upgrade the Client from 17.x to
>>         20.x directly or we need to go through 17.x to 18.x and 19.x
>>         then 20.x
>>
>>         Regards
>>         Navin.
>>
>>
>>
>
>
> -- 
> *Jason L. Simms, Ph.D., M.P.H.*
> Manager of Research and High-Performance Computing
> XSEDE Campus Champion
> Lafayette College
> Information Technology Services
> 710 Sullivan Rd | Easton, PA 18042
> Office: 112 Skillman Library
> p: (610) 330-5632
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.schedmd.com/pipermail/slurm-users/attachments/20201102/8e88af6b/attachment-0001.htm>