0

The situation is similar with [ceph-users] Cannot remove cache tier .

The same as:

The total size and the number of stored objects in the rbd-cache pool oscillate around 5 GB and 3K, respectively, while "rados -p rbd-cache cache-flush-evict-all" is running in a loop. Without it, the size grows to 6 GB and stays there.

And I can't check it from redhad Knowledgebase without subscribe, could anyone help? THX.

Victor Lee
  • 2,467
  • 3
  • 19
  • 37
  • 1
    Can you add more details about your cluster and the workload? Is it possible that the cache-tier is so heavily under load that it can't evict/flush fast enough? Which ceph version is it? Did you follow the steps from the [docs](https://docs.ceph.com/en/latest/rados/operations/cache-tiering/?highlight=cache-tier#removing-a-cache-tier)? – eblock Mar 30 '22 at 10:17
  • My colleague hat a [conversation](https://www.spinics.net/lists/ceph-users/msg41943.html) about removing a cache-tier a few years ago. I can't quite recall all the details, but since cache-tier is not really actively developed and is kind of deprecated, you might need to figure this out by yourself. For example, it might be necessary to set the mode to "forward" although it's not recommended. Maybe you have a test env to try it there first. – eblock Mar 30 '22 at 10:29
  • `ceph -s` get **HEALTH_OK**, the version is 15.2.15 octopus(stable). yes, we follow the steps from that docs. @eblock Thanks for your help, we'll try that. – Victor Lee Mar 30 '22 at 11:50

0 Answers0