1996fanrui commented on code in PR #21960:
URL: https://github.com/apache/flink/pull/21960#discussion_r1110001094
##########
flink-runtime/src/test/java/org/apache/flink/runtime/testutils/MiniClusterResource.java:
##########
@@ -199,6 +200,9 @@ private void startMiniCluster() throws Exception {
new
Configuration(miniClusterResourceConfiguration.getConfiguration());
configuration.setString(
CoreOptions.TMP_DIRS,
temporaryFolder.newFolder().getAbsolutePath());
+ configuration.set(
+ CheckpointingOptions.CHECKPOINTS_DIRECTORY,
+ temporaryFolder.newFolder().toURI().toString());
Review Comment:
FileSystemCheckpointStorage is compatible with more scenarios than
JobManagerCheckpointStorage, for example: scenarios with medium or large state
sizes. But its performance may be a little bit worse in small state scenarios.
I can fix the bug that the test fails, and then we look at the test running
time. If there is no significant increase, `FileSystemCheckpointStorage` can be
used as the default, what do you think?
Actually, I don't think there be much impact on CI running time after
FileSystemCheckpointStorage is the default, because
`JobManagerCheckpointStorage` just works when state size < 5MB, it's small
state. Small state should not consume too much time when using
FileSystemCheckpointStorage.
--
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]
