Live Migration Failed, Multipathd Service Not Running on Destination Host

Problem

Instance live migration fails with below error in ostackhost.log of source host.

ERROR nova.compute.manager [-] [instance: "instance-id"] Pre live migration failed at "destination-host-uuid": RemoteError: Remote error: ProcessExecutionError Unexpected error while running command.Command: multipathd show statusExit code: 1...TRACE nova.compute.manager [instance: "instance-id"] Stderr: u'ux_socket_connect: Connection refused\n'

Environment

• Platform9 Managed OpenStack - All Versions
• multipathd
• pf9-ostackhost

Cause

• The pf9-ostackhos t service on the source and the destination host is configured to use the multipath connection for volumes.

# less /opt/pf9/etc/nova/conf.d/nova_override.conf..volume_use_multipath = True..

• Once the live migration is initiated, a pre-live migration check is performed to confirm if the destination host is running the multipathd service.
• If this pre-live migration check fails, the below command on the destination host will give the error mentioned in the problem section on source host.

$ multipathd show status

• On the destination host, we can observe the error message with the live-migration request ID as below.

ERROR os_brick.initiator.linuxscsi [req-xxxxxxxxxxx user@domain service] multipathd is not running: exit code 1: ProcessExecutionError: Unexpected error while running command.

• Check the status of multipathd service on the destination host.

$ systemctl status multipathd.service

• This issue is most often observed if the multipathd service is not running on the destination host.

Resolution

• Check multipath configuration in / etc/multipath.conf file on destination host.
• If the configuration is correct, start the multipathd service if not running on the destination host.

$ systemctl start multipathd.service