First of all I want to say thanks very much for the SLURM roll - it's
saved me a lot of time and trouble.
I'm using release 6.2-14.11.6 with Rocks 6.2 and am seeing a minor problem
which I think is also present in the 6.2-14.11.7 release.
When my nodes boot up after power saving they often seem to be missing
updates from 411 that happened while they were off. I think these updates
are supposed to be pulled in by /etc/init.d/slurm-prep at boot before
slurm starts. However that script doesn't pull the updates because there's
a syntax error in /etc/rc.d/rocksconfig.d/post-91-slurm . When a node
boots I get an error message from the latter script:
/etc/rc.d/rocksconfig.d/post-91-slurm: line 49: syntax error near
unexpected token fi'
/etc/rc.d/rocksconfig.d/post-91-slurm: line 49:fi'
Because /etc/rc.d/rocksconfig.d/post-91-slurm never completes,
/etc/init.d/slurm-prep always skips fetching updates from 411.
After I made the change in the attached patch to correct the syntax error
my nodes now seem to pick up missed updates when SLURM powers them up.
I hope it's useful,
Probably too late as Werner has already released a new version with the patch, but just in case you don't want to update your SLURM version to 15 right away, what I did was check out the source to the 6.2-14.11.6 roll, apply the patch to the slurm-clients.xml file in that, and build a new roll from that. After I replaced my old SLURM roll with the new one and ran 'rocks create distro' all newly installed nodes had the fix.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hi
First of all I want to say thanks very much for the SLURM roll - it's
saved me a lot of time and trouble.
I'm using release 6.2-14.11.6 with Rocks 6.2 and am seeing a minor problem
which I think is also present in the 6.2-14.11.7 release.
When my nodes boot up after power saving they often seem to be missing
updates from 411 that happened while they were off. I think these updates
are supposed to be pulled in by /etc/init.d/slurm-prep at boot before
slurm starts. However that script doesn't pull the updates because there's
a syntax error in /etc/rc.d/rocksconfig.d/post-91-slurm . When a node
boots I get an error message from the latter script:
/etc/rc.d/rocksconfig.d/post-91-slurm: line 49: syntax error near
unexpected token
fi' /etc/rc.d/rocksconfig.d/post-91-slurm: line 49:
fi'Because /etc/rc.d/rocksconfig.d/post-91-slurm never completes,
/etc/init.d/slurm-prep always skips fetching updates from 411.
After I made the change in the attached patch to correct the syntax error
my nodes now seem to pick up missed updates when SLURM powers them up.
I hope it's useful,
Best wishes
Catherine
Nice catch. Any idea how to make that patch stick to all future kickstarted nodes?
Probably too late as Werner has already released a new version with the patch, but just in case you don't want to update your SLURM version to 15 right away, what I did was check out the source to the 6.2-14.11.6 roll, apply the patch to the slurm-clients.xml file in that, and build a new roll from that. After I replaced my old SLURM roll with the new one and ran 'rocks create distro' all newly installed nodes had the fix.
I have been hoping to avoid learning to (re)build any rolls myself ;-) And on our present delivery schedule I plan to use 15. Thanks.
This bug is only present in this versions:
release-6.1.1-14.11.6
release-6.2-14.11.6
release-6.2-14.11.7
If you are using one of these versions and you don't want or cannot upgrade, I will provide a simple patch(workaround).
Please send me a short e-mail which version you are using.
Best regards
Werner