An Active MDS in Ceph comes across many states during a normal operation. Bobcares, as a part of our Server Management Services, offers solutions to every query that comes our way.

MDS States in Ceph

Throughout typical CephFS operation, the Metadata Server (MDS) transitions between a number of states. Let’s look into those states in this article.

1. up:active

This is the MDS in its typical operational mode. It signifies the availability of the MDS and its rank in the file system.

2. up:standby

This state shows the availability of an MDS to take over for a failed rank. The monitor will automatically perform this action.

3. up:standby_replay

This state is for standby MDS in replay mode. When an active MDS fails, the MDS in this state will replace the failed one.

4. up:boot

During startup, this state transmits to the Ceph monitors. This condition is never present because the Monitor always assigns the MDS to a rank that is open or instructs the MDS to work as a standby.

5. up:creating

The MDS is producing some per-rank metadata and joining the MDS cluster to create a new rank.

6. up:starting

A stopped rank is being restarted by the MDS. It enters the MDS cluster after opening any related per-rank metadata.

7. up:stopping

MDS will accept no new client connections. An active MDS enters to this state as soon as a rank stops.

8. up:replay

MDS replacing a failed rank. The MDS is recovering its journal and other metadata in this state.

9. up:resolve

If there are many active MDS clusters on the Ceph file system, the MDS changes to this state from up:replay.

10. up:reconnect

From up:replay or up:resolve, an MDS moves into this state. The purpose of this state is to encourage client reconnections.

11. up:rejoin

From up:reconnect, the MDS transitions to this state. The MDS is currently re-joining the MDS cluster cache.

12. up:clientreplay

Any client requests that were answered but were not yet durable are being replayed by MDS. The MDS could move into this state from up: rejoin.

13. down:failed

The rank in the file system holds this state.

14. down:damaged

This state is not truly held by an MDS. It is instead applied to the file system’s rank.

15. down:stopped

The file system’s rank holds this state. MDS won’t transition to this state.

[Need further assistance? Get in touch with us.]

Conclusion

The article details the different states an MDS goes through during a normal CephFS operation. Some of the states are very common. While others are transitory states and others belong to failed states.