GeezerVison

When a host has no explicit tags initially, cls.host.hosttags is None.
Passing None to the API does not set the hosttags parameter, so tags
are never cleared during cleanup. Convert None to empty string to properly
clear tags that were set during the test.

Copilot AI left a comment

Zone-wide storage pools (scope=ZONE) are not bound to clusters or pods.
When filtering by clusterid/podid, these pools may be included in results
with clusterid=None and podid=None. Updated tests to:

- test_01: Accept zone-wide pools in clusterid filter results
- test_07: Accept zone-wide pools in podid filter results
- test_12: New test validating zone-wide vs cluster-scoped pools

Both invalid clusterid/podid filters now correctly return zone-wide pools,
and assertions skip None values for zone-wide pools while validating
cluster-scoped pools have their expected scope identifiers.

Validated on two environments:
- Zone-wide only (10.1.112.155): 12/12 tests passed
- Cluster-scoped only (10.0.112.202): 12/12 tests passed

Copilot AI left a comment

TestVolumeEncryption previously skipped entirely on RBD storage pools.
Now that RBD encryption is supported (PR apache#13556), extend it instead of
skipping:

- setUpClass no longer hard-skips RBD pools; records storage_pool_type
  for use by the assertions below.
- check_volume_encryption() branches on pool type: RBD volumes are
  encrypted natively by librbd, so the domain XML carries
  <encryption format='luks2' engine='librbd'>; other pool types keep
  the existing qemu-native <encryption format='luks'> check.
- does_host_with_encryption_support_exists() gained an RBD-specific
  path: the API's "encryptionsupported" field only reflects qemu-native
  LUKS support, so librbd capability is read from the host_details
  table (host.volume.encryption.rbd), joined on the host's uuid.
- get_expected_guest_visible_size() centralizes the existing PowerFlex
  size-overhead adjustment and adds the equivalent for RBD: encrypted
  RBD volumes reserve a 16 MiB LUKS2 header out of the volume's own
  reported size, so the guest sees size - header.
- test_03_root_and_data_volume_encryption now also exercises resize of
  the attached encrypted data volume, reusing its VM instead of
  spinning up a separate one:
  - online (VM stays running): verified via the volume's own API state
    and size, since picking up a live virtio-blk capacity change in the
    guest depends on the guest OS/kernel (the CentOS 5.5 template used
    here predates reliable live-resize notification) and is orthogonal
    to whether CloudStack/libvirt/qemu actually grew the volume.
  - offline (detach, resize, reattach): verified against the guest's
    fdisk output, matched by the expected byte count rather than a
    device letter derived from deviceid, since libvirt is not
    guaranteed to reuse the same target letter after a detach then
    reattach.

Deploy a VM from the direct download template on each type of storage found
(NFS, RBD, SharedMountPoint and local) in a single test, using the storage
pool type as the storage tag.

A disk-only VM snapshot does not contain the guest page cache, so the file
written by the test may be missing after the revert.

Copilot AI left a comment

Copilot AI left a comment

Assert that the encryption format and passphrase secret lists of TestVolumeEncryption are not empty, instead of asserting they are not None, which findall() never returns. Compare the storage pool type case-insensitively, give the direct download test a unique storage tag, skip the live VM migration test for every RBD source pool, and drop an unused variable in test_list_storage_pools.

Copilot AI left a comment