clusterinthecloud / support

If you need help with Cluster in the Cloud, this is the right place
2 stars 0 forks source link

Turn off Hyper-Threading #9

Closed trnh closed 4 years ago

trnh commented 4 years ago

I am trying to find a way to run a post installation script to turn off HT like this

for cpunum in $(cat /sys/devices/system/cpu/cpu*/topology/thread_siblings_list | cut -s -d, -f2- | tr ',' '\n' | sort -un) do echo 0 > /sys/devices/system/cpu/cpu$cpunum/online done

is there anyway to turn HT off.

milliams commented 4 years ago

I'm assuming that you only care about this on the compute nodes, not on the login node?

Currently we only have hooks into customising the image, not into running a post-boot script. That said, it's not a hard change to make. I've added a issue at clusterinthecloud/ansible#83 to track this.

I'm hoping to track down some of these quality-of-life issues over the summer to make these things easier.

For a quick hack, you can add it into the bottom of the script at /home/slurm/bootstrap.sh which should be run on all newly-booted nodes.

The only potential issue would be that Slurm expects a certain number of cores per instance type and so there might need to be a tweak elsewhere too.

trnh commented 4 years ago

bootstrap.sh solution should work fine , the reason to turn off HT on AWS for MPI application to use physcal core.

Thank you