| 1 |
--- |
| 2 |
issues: |
| 3 |
- Cinder services are now automatically downgrading RPC |
| 4 |
messages to be understood by the oldest version of a |
| 5 |
service among all the deployment. Disabled and dead |
| 6 |
services are also taken into account. It is important to |
| 7 |
keep service list up to date, without old, unused |
| 8 |
records. This can be done using ``cinder-manage service |
| 9 |
remove`` command. Once situation is cleaned up services |
| 10 |
should be either restarted or ``SIGHUP`` signal should |
| 11 |
be issued to their processes to force them to reload |
| 12 |
version pins. Please note that cinder-api does not |
| 13 |
support ``SIGHUP`` signal. |
| 14 |
upgrade: |
| 15 |
- If during a *live* upgrade from Liberty a backup service |
| 16 |
will be killed while processing a restore request it may |
| 17 |
happen that such backup status won't be automatically |
| 18 |
cleaned up on the service restart. Such orphaned backups |
| 19 |
need to be cleaned up manually. |
| 20 |
- When performing a *live* upgrade from Liberty it may |
| 21 |
happen that retype calls will reserve additional quota. |
| 22 |
As by default quota reservations are invalidated after |
| 23 |
24 hours (config option ``reservation_expire=86400``), |
| 24 |
we recommend either decreasing that time or watching for |
| 25 |
unused quota reservations manually during the upgrade |
| 26 |
process. |