lop-devops/tests

Sprint 1: test bz: CPU & Memory hotplug: DRC state info transmission for migration

bssrikanth opened this issue · 5 comments

bssrikanth commented

This is test bz to track the testcases development for below new feature targeted for sprint 1:
123907 CPU & Memory hotplug: DRC state info transmission for migration

@sathnaga will be test owner.

@sathnaga please update with the testcases meant for testing this new feature.

cde:info Mirrored with LTC bug #158309 </cde:info>

bssrikanth commented

I believe testing this feature is similar to #60. So same testcase should hold good.

  1. Create guest with shared storage in NFS
  2. Enabled ports 49152:49216 in iptables, virt_use_nfs -> on
  3. Mount the image location in destination and started migration.
  4. Perform CPU Hotplug and Hotunplug, migrate to other host and perform CPU Hotplug and Hotunplug
    CPU hot unplug should be successful
bssrikanth commented

since its similar to #60 Bala would be testing this feature.

cdeadmin commented

------- Comment From danielhb@br.ibm.com 2017-09-06 07:45:02 EDT-------
In step 4:

"4. Perform CPU Hotplug and Hotunplug, migrate to other host and perform CPU Hotplug and Hotunplug"

I am unsure if this is what it meant but If you hotplug and unplug a CPU before migration, migrate, then hotplug and unplug a CPU in destination you will not be testing the bug reported in 123907.

The sequence of events to reproduce that issue is:

  • start VM in source host
  • hotplug a CPU
  • migrate to destination
  • hotunplug the CPU

Or, putting in a sentence:

"4. Perform CPU Hotplug, migrate to other host and perform CPU Hotunplug"

Thanks,

Daniel

cdeadmin commented

------- Comment From bssrikanth@in.ibm.com 2017-09-08 01:10:55 EDT-------
(In reply to comment #5)
> In step 4:
>
> "4. Perform CPU Hotplug and Hotunplug, migrate to other host and perform CPU
> Hotplug and Hotunplug"
>
> I am unsure if this is what it meant but If you hotplug and unplug a CPU
> before migration, migrate, then hotplug and unplug a CPU in destination you
> will not be testing the bug reported in 123907.
>
> The sequence of events to reproduce that issue is:
>
> - start VM in source host
> - hotplug a CPU
> - migrate to destination
> - hotunplug the CPU
>
> Or, putting in a sentence:
>
> "4. Perform CPU Hotplug, migrate to other host and perform CPU Hotunplug"
>
>
> Thanks,
>
>
> Daniel

@daniel thanks for the review. Ack'ed on behalf of @bala.

bssrikanth commented

This is tested.

  1. Hotplug 64 cpus to guest in source.
    virsh setvcpus avocado-vt-vm1-migration 64 --live
  2. migrate vm to destination machine.
    virsh migrate avocado-vt-vm1-migration qemu+ssh://9.40.193.221/system --live --verbose
    root@9.40.193.221's password:
    Migration: [100 %]
  3. Hotunplug 56 cpus to guest in destination
    virsh setvcpus avocado-vt-vm1-migration 8 --live
    echo $?
    0
    Checked the respective change in cpu nos from inside guest. It's successful.