Verify that VMs stay up when disks are evicted

  • Related issues
    • #1334 Volumes fail with Scheduling Failure after evicting disc on multi-disc node
    • #5307 Replicas should be evicted and rescheduled to other disks before removing extra disk

Verification Steps

  1. Created 3 nodes Harvester.

  2. Added formatted disk (called disk A) to node0 VM in the harvester node page.

  3. Added disk tag test on following disk in the longhorn page.

    1. disk A of node0
    2. root disk of node1
    3. root disk of node2

  4. Created storage class with disk tag test and replica 3.

  5. Created volume (called B) with previous storage class. You should check scheduling status in the longhorn page.

  6. Created VM with volume B as extra volume (not boot volume).

  7. The node scheduling status should look like this. images/volumes/1334-image-01.png

  8. Delete disk A in the harvester node page and revisit it, error message should be displayed.

  9. Check Eviction Requested is True and Scheduling is Disable on disk A in the longhorn dashboard. images/volumes/1334-image-02.png

  10. Add disk tag test to root disk of node0.

  11. Now disk A should be removed since replica condition is satisfied again. images/volumes/1334-image-03.png

Expected Results

  1. Disk A is removed on the harvester node page.
  2. VM is running in whole steps.