Bug 1342681

Description of problem:

After upgrading to 3.6.6, a live merge of a VM with 4 disks failed, but only on the engine.

Snapshots for two of the disks were successfully merged, and the other two failed due to a "network connectivity issue".

These merges involved the active layer.

Prior to the merges, all four disks consisted of two volumes.

On the host though, all four merges completed.

This left the following inconsistencies;

1) RHEV database;

- Two disks still had two images, with the base images being marked 'illegal' (imagestatus = 4).

- The original active image for these two disks were still marked as 'active'.


2) Storage;

- The two volumes above still physically existed in the storage domain, but were no longer 'open' (just 'active').

- The volume metadata for these volumes were marked 'ILLEGAL'.



Question: Rather than manually modifying the database and the metadata and removing volumes, would retrying the merges have been successful in this case ?  That is, since the active volumes in the database for these two disks were actually no longer 'open' on the host, would a retry have resulted in these just being removed and the above inconsistencies being reconciled ?  


Version-Release number of selected component (if applicable):

RHEV 3.6.6


How reproducible:



Steps to Reproduce:
1.
2.
3.

Actual results:



Expected results:


Additional info:

I