Skip to content

Conversation

jovial
Copy link
Contributor

@jovial jovial commented Feb 18, 2021

Using total memory as value of RealMemory in slurm.conf does not allow for
OS overheads and can cause slurm to srain the nodes with: LowRealMemory.

Fixes #92.

Using total memory as value of `RealMemory` in slurm.conf does not allow for
OS overheads and can cause slurm to srain the nodes with: `LowRealMemory`.

Fixes #92.
@jovial jovial requested a review from sjpb February 18, 2021 12:52
Copy link
Collaborator

@sjpb sjpb left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

1x minor nit re. docs, templating looks good

@sjpb
Copy link
Collaborator

sjpb commented Feb 25, 2021

@sjpb to run thro on labs openhpc_tests, and merge if OK.

@sjpb
Copy link
Collaborator

sjpb commented Feb 25, 2021

Testing:

[centos@*-hpc-0 ~]$ slurmd -C
NodeName=*-hpc-0 CPUs=16 Boards=1 SocketsPerBoard=2 CoresPerSocket=8 ThreadsPerCore=1 RealMemory=128616
UpTime=1-05:31:27
[centos@*-control ~]$ grep RealMemory /etc/slurm/slurm.conf
    RealMemory=122185 \
    RealMemory=122185 \

So can see default 0.95 multiplier has been applied.

Now running:

ansible-playbook ansible/adhoc/test.yml -e "openhpc_tests_nodes=*-hpc-[0-3]"

...

@sjpb
Copy link
Collaborator

sjpb commented Feb 25, 2021

Ran openhpc_tests ok, nodes still up afterwards so lets merge this @jovial.

@jovial jovial merged commit cb5a80a into master Feb 26, 2021
@jovial jovial deleted the fix/#92 branch February 26, 2021 10:03
@jovial
Copy link
Contributor Author

jovial commented Feb 26, 2021

Ran openhpc_tests ok, nodes still up afterwards so lets merge this @jovial.

Your wish is my command - thanks for testing :)

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

Using 100% of real memory as default for ram_mb is too high
2 participants