Search results

I spoke too soon. My crashing is still occurring on 6.2.16-14-pve. It's probably unrelated to your issue. Sorry to pollute your thread with unnecessary information.

After pinning the kernels for my cluster to 6.2.16-14-pve, and updating them all to have the same package versions, the 3rd node is has not crashed in a few days. I am not 100% sure which action resolved the issue, getting off kernel 6.2.16-15-pve, or updating all the nodes to have the same...

I have a cluster of 3 physically identical nodes. One of the 3 is randomly crashing overnight. After looking at pveversion, it looks like the one that is crashing was updated while the other 2 were not updated (not sure how I managed to do that, but I must have). The crashing one is using...

My TR 3975WX is used for 4 gaming VMs and 6 other server VMs. I usually use this server in UMA mode, not NUMA. I switch to NUMA for the above tests. I pin each gaming VM each to their own set of CCDs. Each core has it's own L1-L2 cache and each CCD has it's own 16MB L3 cache along with 4 cores...

So, I ran several tests. TLDR: As far as this testing of CPU pinning goes, numactl can do most of what taskset can do and more. One of those additional features offers better numa memory latency but is a double edged sword for servers that are numa OOM. If you only want to pin a vm to cores in...

For the better part of a year and a half I have been using a variation of the hookscript above. Each VM is setup with the hookscript and a separate file that defines it's CPU set. This was really annoying so, I wrote a patch and replied to the bug tracker which adds the affinity value to the...