Possible for cinder-volumes partition size to exceed physical disk size on disk replacement

[Maria Yousaf]

Brief Description
-----------------
Partition size on AIO-SX cinder disk replacement procedure can be larger than physical disk due to lack of validation. This will result in errored partition that cannot be deleted.

Severity
--------
Critical

Steps to Reproduce
------------------
1) Create cinder volume(s) and cinder snapshot(s). Backup and store volume data.
2) Make note of serial # of cinder disk to be replaced and disk size. In this case, disk being replaced was 480 GB.
3) Lock controller
4) Power down controller
5) Replace disk with disk of larger size. In this case, disk was 800 GB.
6) Power on controller
7) Ensure new disk serial # is listed
8) Cinder has an in-use partition. Use horizon to modify partition to larger size. In this case, partition was accidentally modified to larger than the available_gib size (745.209). Partition size was set to 800 gib.
9) This results in partition going into internal error script state (Error: Internal Script Error state) eventually
10) Partition cannot be deleted either via the CLI or horizon. LVG and PVG cannot be deleted either.
11) Attempted workaround was to reboot AIO-SX node. This puts the partition of 800 gib into In-use state as opposed to error however node will not become available.

Expected Behavior
------------------
Partition creation larger than physical disk will be rejected.

Actual Behavior
----------------
Partition size creation can be larger than physical disk.

Reproducibility
---------------
Tried once.

System Configuration
--------------------
One node system

Branch/Pull Time/Commit
-----------------------
2018-10-02_20-18-00

[Ghada Khalil (gkhalil)]

stx.2019.03 - specific to disk replacement scenarios

[Bruce Jones (brucej)]

Cindy can you please get with Vivian and re-assign this to someone on her team, thanks!

[Maria Yousaf (myousaf)]

Changing state to invalid, due to the introduction of ceph on all StarlingX configurations. Disk replacements will need to be tested in general, as part of regression testing prior to release.

[Liang Fang (liangfang)]

@Maria "Invalid" is the Status of "Closed", right? From your comment, this test case is not valid any more when ceph introduced on all configurations. As the bug is invalid, does it make sense to remove the tag stx.2.0 and assign the bug back to yourself? In my understanding, stx.2.0 is a gating tag that would block release.

[Cindy Xie (xxie1)]

@liangfang "invalid" is equivalent to "closed". It's OK to leave it to anybody as none of the queries will run it for now. The "stx.2.0" query will not include the bugs in "invalid" state.