Target host in nova DB got updated to new compute while migration failed
Bug #1653718 reported by
Sahid Orentino
This bug affects 3 people
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
OpenStack Compute (nova) |
In Progress
|
Undecided
|
Sahid Orentino |
Bug Description
During live-migration if the process goes in a unpredictable situation for example if QEMU suffers an issue Nova could still consider the migration has succeeded even if not.
The VM on source node can be still registered even if stopped. In the worst scenario, the operator could start it and so two VMs running could share the same disk.
We should fix that issue, do not consider a migration to have succeeded if it was not the case. For that we should handle the return migrateToURI*
[1] http://
[2] http://
Changed in nova: | |
assignee: | nobody → sahid (sahid-ferdjaoui) |
status: | New → In Progress |
To post a comment you must log in.
Driving any logic off the return value of migrateToURI* is really a bad idea. When migrateToURI* fails, there are many different states the VM can be in on either the src/dst host, and the migrateToURI return status gives no useful info about which applies. Migration might even have 100% succeeded, despite migrateToURI returning an error. This is why Nova stopped looking at migrateToURI error status and switched to the job info records, to solve https:/ /bugs.launchpad .net/nova/ +bug/1414065