virsh nodecpumap --pretty shows unwanted cpu map output
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
libvirt (Ubuntu) |
Fix Released
|
Medium
|
Taco Screen team | ||
Xenial |
Fix Released
|
Medium
|
Unassigned |
Bug Description
[Impact]
* virsh nodecpumap --pretty --pretty shows wrong result on machine with
many cores
* Since admin and scripts might rely on those they should be fixed.
Not only are they not nice, they are depending on the case even wrong
sometimes.
* an example is like:
virsh nodecpumap --pretty
CPUs present: 120
CPUs online: 120
CPU map: 0-119,192,
[Test Case]
* Go to a huge machine with many CPUs (a virtual machine with faked vcpus
might help, but stay aware of issues with 2nd level virtualization)
* run "virsh nodecpumap --pretty"
* Tests with extra splits into numa nodes can be done, but are not
necessary to show the effect
[Regression Potential]
* Fix is upstream quite while now without further changes which should make it safe.
* Also since the code is restricted to "virsh nodecpumap" and not reused
any potential regression should be limited to that sub-command which is
nice for some use-cases but not a core function in general.
[Other Info]
* n/a
----- original report ------
Please cherry pick this issue.
https:/
commit 3a703b8120d9f99
Changed in libvirt (Ubuntu): | |
importance: | Undecided → Medium |
status: | New → Triaged |
description: | updated |
Default Comment by Bridge