Red Hat Bugzilla – Bug 1313882
nova - multiple block_device_mapping records
Last modified: 2018-05-21 21:53:25 EDT
This bug seems like a likely duplicate of BZ 1305408, but will not mark it as such yet before we are sure.
We will need to get the logs to investigate further, preferably from the time when the stress testing happened that leads to the bug.
Additionally - we would also want _exact_ commands used to do the stress testing as opposed to just descriptions.
Ok so I spent some time digging into the already reported bugs and it seems that this has been reported already:
The proposed patch:
seems outdated though, so I'll try to work on a more up-to-date one.
Opening this bug up so that other duplicates can be closed out.
*** Bug 1342347 has been marked as a duplicate of this bug. ***
*** Bug 1341937 has been marked as a duplicate of this bug. ***
*** Bug 1340996 has been marked as a duplicate of this bug. ***
*** Bug 1305408 has been marked as a duplicate of this bug. ***
*** Bug 1376818 has been marked as a duplicate of this bug. ***
Moving back to ON_DEV, the current plan is to remove duplicate BDMs in 14/R with an online migration before then introducing a new constraint around volume_id and instance_uuid on the table in 15/S. While this doesn't fix the underlying race it will ensure duplicate records are not left in the DB.