cinder-volume fails to initialize properly

Bug #1595735 reported by Eugene Nikanorov
6
This bug affects 1 person
Affects Status Importance Assigned to Milestone
Mirantis OpenStack
Status tracked in 10.0.x
10.0.x
Invalid
High
Eugene Nikanorov
7.0.x
Invalid
High
MOS Maintenance
8.0.x
Invalid
High
MOS Maintenance
9.x
Invalid
High
Eugene Nikanorov

Bug Description

The following output from pvs command prevents cinder-volume with LVM driver to initialize properly.
A stack trace could be seen on cinder-volume startup and on any create-volume operation.

root@node-33:~# pvs --noheadings --unit=g
  Couldn't find device with uuid fZPJvz-w1sb-Gj4T-aJiG-1pse-v0Ur-VwKzoa.
  /dev/sda3 cinder lvm2 a-- 2793.91g 2528.91g
  /dev/sdb3 cinder lvm2 a-- 2793.91g 2793.91g
  /dev/sdc3 cinder lvm2 a-- 2793.91g 2793.91g
  /dev/sdd3 cinder lvm2 a-- 2793.91g 2793.91g
  /dev/sde4 os lvm2 a-- 222.96g 0.00g
  unknown device cinder lvm2 a-m 11175.26g 11175.26g

The problem could be worked around by running:
vgreduce --clearmissing cinder

Expected result:
On driver initialization failure cinder-volume should exit to clearly indicate that it is not operational.

Revision history for this message
Eugene Nikanorov (enikanorov) wrote :

This was discovered on MOS 7 (Kilo), but the LVM driver code is the same in 8, 9 and in upstream cinder, so it makes sense to file it upstream as well.

Changed in mos:
assignee: nobody → MOS Cinder (mos-cinder)
tags: added: need-info
tags: removed: need-info
tags: added: need-info
Revision history for this message
Ivan Kolodyazhny (e0ne) wrote :

Eugene, do you have any logs and/or steps to reproduce this issue?

Revision history for this message
Eugene Nikanorov (enikanorov) wrote :
Revision history for this message
Ivan Kolodyazhny (e0ne) wrote :

According to logs, there is something went wrong with LVM. According Cinder is able to run 'vgs' command but fails to create using 'lvcreate' command.

"Expected result:
On driver initialization failure cinder-volume should exit to clearly indicate that it is not operational." - it will be done in scope of [1] feature

"but the LVM driver code is the same in 8, 9 and in upstream cinder, so it makes sense to file it upstream as well." - it's totally wrong to compare code only in driver. We should to compare manager's and scheduler's code too

[1] https://github.com/openstack/cinder-specs/blob/master/specs/newton/summarymessage.rst

To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.