[RBD -> SolidFire] Cache mode was not updated after volume retype

Bug #1999252 reported by mikhail
6
This bug affects 1 person
Affects Status Importance Assigned to Milestone
Cinder
Incomplete
High
Unassigned

Bug Description

Steps to reproduce
1. Set nova.conf
[libvirt]
disk_cachemodes = network=writeback

2. Create RBD volume and attach it to an instance as additional volume.

    <disk type='network' device='disk'>
      <driver name='qemu' type='raw' cache='writeback' discard='unmap'/>
      <auth username='cinder'>
        <secret type='ceph' uuid='<secret uuid>'/>
      </auth>
      <source protocol='rbd' name='volumes/volume-<volume id>'>
        <host name='192.168.1.37' port='6789'/>
        <host name='192.168.1.38' port='6789'/>
        <host name='192.168.1.55' port='6789'/>
      </source>
      <mirror type='block' job='copy'>
        <format type='raw'/>
        <source dev='/dev/sdd'/>
        <backingStore/>
      </mirror>
      <target dev='vdg' bus='virtio'/>
      <serial>5a71c185-1a17-48da-9d70-c73cdbf73681</serial>
      <alias name='virtio-disk6'/>
      <address type='pci' domain='0x0000' bus='0x00' slot='0x0b' function='0x0'/>
    </disk>

3. Retype volume in in-use state to SolidFire or NetApp ONTAP backend. Retype mode is On-demand.

4. Volume was moved to another backend, but the cache mode was not changed

    <disk type='block' device='disk'>
      <driver name='qemu' type='raw' cache='writeback' discard='unmap'/>
      <source dev='/dev/sdd'/>
      <backingStore/>
      <target dev='vdg' bus='virtio'/>
      <serial>5a71c185-1a17-48da-9d70-c73cdbf73681</serial>
      <alias name='virtio-disk6'/>
      <address type='pci' domain='0x0000' bus='0x00' slot='0x0b' function='0x0'/>
    </disk>

In the end of those steps we are not able to live-migrate this instance due to the following error:

Live Migration failure: Unsafe migration: Migration may lead to data corruption if disks use cache != none or cache != directsync: libvirt.libvirtError: Unsafe migration: Migration may lead to data corruption if disks use cache != none or cache != directsync

mikhail (kraynovmv)
summary: - [RBD -> SolidFire] Cache mode was not updated after retyping volume
+ [RBD -> SolidFire] Cache mode was not updated after volume retype
Changed in cinder:
importance: Undecided → High
Revision history for this message
Sofia Enriquez (lsofia-enriquez) wrote :

Hi mikhail (kraynovmv), hope this message finds you well.
Please share the cinder version you are referring to.
Is the error message reported by c-vol or by nova side?
Thanks,
Sofia

tags: added: cache live-migration
Changed in cinder:
status: New → Incomplete
Revision history for this message
mikhail (kraynovmv) wrote :

Hello Sofia,
I'm using cinder from Train version - 15.4.0
The error comes from nova side.

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.