Description of problem: keystone-manage token_flush isn't running frequently enough and will break on some occasion thus accumulating all tokens until a table drop is done because cleaning up those tokens isn't possible any longer. It should be run hourly or even more frequently. Version-Release number of selected component (if applicable): How reproducible: Always as soon as you have a bit of activity in an environment, daily cleanup will fail Steps to Reproduce: 1. Get much activity on the cluster 2. Wait a bit until the daily token flush fails 3. Actual results: Cleanup fails Expected results: Cleanup never fails Additional info: I've seen this on RHOSP 10 and RHOSP 9. Perhaps 9.
This will make recovering galera database corruption through pacemaker very hard because it will time out while trying to sync 160GB (in this current case).