Incorrect VM IOPS metrics

[SviridoffA]

problem

VM metrics show incorrect IOPS. This may be related to the type of my primary storage. I’m using Linstor. When we perform a fio test, we get the following results:

But VM metrics shows in this time:

versions

ACS 4.19.3
Primary storage - Linstor 1.31.3
OS: Ubuntu 24.04.2

The steps to reproduce the bug

Just perform a fio test and check the metrics.

What to do about it?

It would be great if this metric could be fixed.

[UAnton]

Same promlem...

[DaanHoogland]

Are you also using Linstor @UAnton ?
@rp- is this a known issue?

[UAnton]

@DaanHoogland hi,
No, I'm was try on NFS when deploy test cluster and get same issue

[rp-]

Are you also using Linstor @UAnton ? @rp- is this a known issue?

I don't think this is related to Linstor, as I think these IOP numbers come from libvirt, as I don't recall implementing any IOP statistics.

[SviridoffA]

Not sure this related, but just as additional information, my management servers has a non-En locale. I wanted to mention this because we’ve CPU metric issues related to locale in the past.

[DaanHoogland]

ok, so this is not storage type related. What does the graph below this one say, @SviridoffA ? the one that has reads and writes seperated in different values. I am wondering if this is a stats collector issue, a UI issue or a historic data/accumulation issue.

[SviridoffA]

@DaanHoogland
Results, including the Read & Write graph, are below. Looks like incorrect too.

[DaanHoogland]

I’ve seen people show behaviour like this before.

• Can we reproduce easily ?
• is it for all VMs or just a few?
• is it just kvm, or even just ubuntu ?
  I think this one will be hard to pinpoint.