Used hugepage of compute node does not include the mem which is used by system process
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
OpenStack Compute (nova) |
New
|
Undecided
|
Unassigned |
Bug Description
Description
===========
There are some system processes which use hugepage when compute node starts. But when we report numa_topology resources, we do not consider these. And we do not consider it by reserving hugepage.
Expected result
===============
Used NUMAPagesTopology information in compute node is all system used hugepage. Or we can consider it by reserving hugepage mem.
Actual result
=============
Used hugepage information in compute node is not correct. It can cause select a host which can not build a instance using hugepage.
Environment
===========
Os which supports KVM and hugepage. And we build instance which uses hugepage.
Logs & Configs
==============
Flavor include extra_specs hw:mem_
Thanks for the bug report, can you add more details about the version you have seen this issue?