Multiple physical drives for nova-local are not supported, resulting in compute node failing to unlock
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
StarlingX |
Fix Released
|
Medium
|
Erickson Silva de Oliveira |
Bug Description
Brief Description
-----------------
The procedure is to add nova-local ephemeral storage to the compute nodes. For example:
system host-lvg-add $NODE nova-local
system host-disk-list $NODE | awk '/\/dev\
If more than a single physical disk is added, such as calling:
system host-lvg-add $NODE nova-local
system host-disk-list $NODE | awk '/\/dev\
system host-disk-list $NODE | awk '/\/dev\
system host-disk-list $NODE | awk '/\/dev\
The compute node will fail on unlock with and error similar to:
2024-04-
Severity
--------
Provide the severity of the defect.
Major: System/Feature is usable but degraded to a single physical disk.
Steps to Reproduce
------------------
Add two or more physical drives to nova-local on a compute node
system host-label-assign $NODE openstack-
system host-lvg-add $NODE nova-local
system host-disk-list $NODE | awk '/\/dev\
system host-disk-list $NODE | awk '/\/dev\
system host-disk-list $NODE | awk '/\/dev\
system host-unlock $NODE
Note, this was tested on a compute node with standard controllers.
Expected Behavior
------------------
nova-local created successfully
Actual Behavior
----------------
Failed to create nova-local
Reproducibility
---------------
Reproducible
System Configuration
-------
Standard system
Timestamp/Logs
--------------
2024-04-
Workaround
----------
One may add one physical disk to nova-local per lock/unlock. Meaning if one wishes to add three physical disks to nova-local, one can do so by locking, adding a disk, unlock. Then wait for it to be online, lock again and add a second driver.
Changed in starlingx: | |
status: | New → In Progress |
tags: | added: stx.storage |
Changed in starlingx: | |
assignee: | nobody → Erickson Silva de Oliveira (esilvade) |
Changed in starlingx: | |
importance: | Undecided → Medium |
tags: | added: stx.10.0 |
Reviewed: https:/ /review. opendev. org/c/starlingx /stx-puppet/ +/915438 /opendev. org/starlingx/ stx-puppet/ commit/ 349f4e97999bc9e 7561c0a6bbc7751 79343f8452
Committed: https:/
Submitter: "Zuul (22348)"
Branch: master
commit 349f4e97999bc9e 7561c0a6bbc7751 79343f8452
Author: Hediberto C Silva <email address hidden>
Date: Wed Apr 10 14:18:25 2024 -0300
Fix puppet class to wipe new PV
With the change made in [1], when processing more than one
nova-local, puppet fails with "Duplicate declaration", due
to the exec "vgchange -an nova-local".
To resolve this, the variable $name was added, so that it
becomes dynamic.
Furthermore, the review mentioned above deactivate VG to perform
wiping, however, it is not activated again after that. So to
resolve this, another exec was added, so that VG is activated
after wipe new PG.
[1]: https:/ /review. opendev. org/c/starlingx /stx-puppet/ +/863871
Test Plan:
SX: Delete instances fs and add 4 nova-local
B&R with nova-local instead of instances
DX+: Add 4 nova-local in compute-0
B&R with nova-local created
STD: Add 4 nova-local in compute-0
B&R with nova-local created
Closes-Bug: 2061526
Change-Id: I7449c5cd719954 1551dccee17e22a 8bda48414e1
Signed-off-by: Hediberto C Silva <email address hidden>
Signed-off-by: Erickson Silva de Oliveira <email address hidden>