Slow monitoring of the live migrating VMs on destination host

[vholer]

Description
If large memory VM is being live migrated, the monitoring on destination host already polls the (paused) VM over libvirt for block device statistics. These queries can take an extremely long time, resulting in the monitoring timeouts. Also, gathering such statistics for the migrated VM on destination host doesn't make sense since the VM is still running on the source host.

  24056 pts/1    Sl+    0:00  |                       |                       \_ ruby ../../vmm/kvm/poll -t
  24080 pts/1    S+     0:00  |                       |                           \_ sh -c virsh --connect qemu:///system --readonly domblkstat one-20 vda 2>/dev/null
  24081 pts/1    Sl+    0:00  |                       |                               \_ virsh --connect qemu:///system --readonly domblkstat one-20 vda
...
  23914 pts/1    Sl+    0:00  |                                       \_ ruby ../../vmm/kvm/poll -t
  24128 pts/1    S+     0:00  |                                           \_ sh -c virsh --connect qemu:///system --readonly domblkstat one-20 hda 2>/dev/null
  24129 pts/1    Sl+    0:00  |                                               \_ virsh --connect qemu:///system --readonly domblkstat one-20 hda

To Reproduce

1. have large memory VM
2. run CPU / memory intensive application inside
3. live migrate on different HV host
4. experience monitoring timeouts, stale domblkstat

Details

• Affected Component: Storage
• Hypervisor: KVM
• Version: 5.6.0

Progress Status

• Branch created
• Code committed to development branch
• Testing - QA
• Documentation
• Release notes - resolved issues, compatibility, known issues
• Code committed to upstream release/hotfix branches
• Documentation committed to upstream release/hotfix branches

[vholer]

PRs to merge:

• B #2388: Slow monitoring of the live migrating VMs on destination host #2401 into one-5.6 and master
• B #2388: Update resolved issues for 5.6.1 docs#360 into one-5.6-maintenance