nova-compute is currently suffering from race conditions when libvirtd
is being restarted during the startup of nova-compute. (For details see:
https://bugs.launchpad.net/nova/+bug/1654207
Our recipes currently are implemented in a way that makes that race
condition trigger pretty often on the initial deploymented and
especially on the crowbar_join run during the upgrade from an older
release. One reason for this is, that the libvirtd restart happens
after/during the restart of nova-compute.
This patch adds a workaround so that libvirtd is restarted before
nova-compute, by utilizing immediate notifications.
(cherry picked from commit 8c44da82f835c81e2377e38be5e1146a456e532d)
(cherry picked from commit fba82126ad288d0178c9b51a2d3c3cbdbd7b2eb5)
nova-compute is currently suffering from race conditions when libvirtd is being restarted during the startup of nova-compute. (For details see: https://bugs.launchpad.net/nova/+bug/1654207
Our recipes currently are implemented in a way that makes that race condition trigger pretty often on the initial deploymented and especially on the crowbar_join run during the upgrade from an older release. One reason for this is, that the libvirtd restart happens after/during the restart of nova-compute.
This patch adds a workaround so that libvirtd is restarted before nova-compute, by utilizing immediate notifications.
(cherry picked from commit 8c44da82f835c81e2377e38be5e1146a456e532d) (cherry picked from commit fba82126ad288d0178c9b51a2d3c3cbdbd7b2eb5)