pnowojski commented on a change in pull request #18086:
URL: https://github.com/apache/flink/pull/18086#discussion_r768806302
##########
File path:
flink-runtime/src/main/java/org/apache/flink/runtime/checkpoint/CheckpointOptions.java
##########
@@ -179,6 +254,16 @@ public CheckpointStorageLocationReference
getTargetLocation() {
return targetLocation;
}
+ /**
+ * Returns whether the snapshot needs to be a full one in the sense if it
can share artefacts
+ * with prior snapshots.
+ *
+ * @return If it is forced to be a full snapshot it should not share any
artefacts.
+ */
+ public boolean isForcedFullSnapshot() {
+ return forcedFullSnapshot;
+ }
Review comment:
In the context of https://issues.apache.org/jira/browse/FLINK-25276 I've
realised that "force full snapshot" will not be enough. For incremental
savepoints we will actually need a yet another different operating mode. In
total we would have:
1. (current master) you can re-use the artefacts as you wish
2. "forceFullSnapshot"/"no-claim" mode, you can not re-use old artefacts,
but the newly uploaded artefacts in this snapshot can be re-used in the future
3. "incremental savepoint" mode, you can not re-use old artefacts and newly
uploaded artefacts can not be re-used in the future
So I would propose to change this `boolean` flag to `enum`. In this PR this
enum would have two values (1. and 2.). For FLINK-25192, we would add the 3rd
mode. I would consider putting this enum inside `CheckpointType`?
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
