Masakari creates two notifications on host failure

[yosef]

When a host failure happens, two same notifications with different uuid are created, I am on wallaby, having 3 controllers, deployed by kolla, and this error is not happening on stage configuration, but happening on production.

When two notifications are picked on the 3 node engines, they are parallely executed and cause huge disturbance, I have tried to prevent this as a pre recovery task, but there is no notification uuid that i can access there. I could try to run a custom lock in the pre task, locking this host on redis for whom have arrived sooner, and the later onces exit, but I think there can be a better solution.

I would appretiate any help and guidance to finding a workaround.

Also I can provide more information if requested.