Siyao Meng created HDDS-7744:
--------------------------------
Summary: [Snapshot] Implement new quota including keys exclusively
used in snapshots
Key: HDDS-7744
URL: https://issues.apache.org/jira/browse/HDDS-7744
Project: Apache Ozone
Issue Type: Sub-task
Reporter: Siyao Meng
(Excerpt from the design doc)
g) Quota considerations:
1. Currently, quota usage is [immediately
decreased|https://github.com/apache/ozone/blob/ddbe71d3f90ae14ecf29f51c2265f1f9c3171668/hadoop-ozone/ozone-manager/src/main/java/org/apache/hadoop/ozone/om/request/key/OMKeyDeleteRequest.java#L166-L168]
when a key is successfully confirmed deleted in OM, rather than when its key
blocks are reclaimed. As a result, if we do not change the logic in
validateAndUpdateCache in OMKeyDeleteRequest (and the FSO equivalent) to make
it aware of snapshots, effectively those key space exclusively used in
snapshots won’t count against the quota, because their quota usage is
“released” the minute the keys are deleted from the active FS.
2. On one hand, it sounds reasonable to leave the current quota handling logic
in the active file system as-is, because in a case where a regular user Bob has
exhausted its bucket quota and at this point an admin Alice takes a snapshot on
the bucket, Bob would have no way to release his bucket quota usage if snapshot
usage also counts against the current Ozone quota implementation. Only admins
are able to delete and create snapshots.
3. If needed, we could introduce another kind of quota which includes snapshot
usage in the future, [similar to what ZFS
has|https://docs.oracle.com/cd/E53394_01/html/E54801/gazvb.html]. ZFS has
refquota which doesn't count snapshot usage (like what we currently have in
Ozone), and quota which counts snapshot usage (and others) as well.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
