2015-10-20 23:40:29 |
John Griffith |
description |
sometimes we check/set quotas in cinder.volume.api before the schedule call; other times we check it on the other side of the call in cinder.volume.manager
We should probably be consistent with these, and we should probably do these on the API side so we can return a reasonable response to the caller before launching the rpc call. |
sometimes we check/set quotas in cinder.volume.api before the schedule call; other times we check it on the other side of the call in cinder.volume.manager
We should probably be consistent with these, and we should probably do these on the API side so we can return a reasonable response to the caller before launching the rpc call.
This is mostly due to blocking vs non-blocking calls but we should take a look at cleaning some of these up. We should be able to always check the reservations in volume.api and then roll-back adjust in the case of failure in the volume.manager if need be. |
|