collectd should report the physical CPU count for compute nodes instead of virtual CPU
Bug #1590315 reported by
Simon Pasquier
This bug affects 1 person
| Affects | Status | Importance | Assigned to | Milestone | |
|---|---|---|---|---|---|
| StackLight |
Triaged
|
Low
|
LMA-Toolchain Fuel Plugins | ||
Bug Description
The collectd plugin collecting hypervisor statistics multiplies the 'vcpus' value by the assumed allocation ratio. This was done in this way to be consistent with the vcpus_used' value for which the allocation ratio is already applied.
However if the operator changes the default allocation ratio (1:8 for MOS) or uses aggregates to customize the allocation ratios, the computed value is incorrect. So we'd better off leave the collected value unchanged and multiply this value by the allocation ratio in Grafana so that the user can easily change it.
Revision history for this message
| Simon Pasquier (simon-pasquier) wrote | #1 |
| Changed in lma-toolchain: | |
| milestone: | none → 0.10.1 |
| Changed in lma-toolchain: | |
| milestone: | 0.10.1 → 0.11.0 |
| Changed in lma-toolchain: | |
| milestone: | 0.11.0 → none |
To post a comment you must log in.
For now, a user can modify the allocation ratio that collectd applies by changing the CpuAllocationRatio parameter in /etc/collectd/