RocMarshal commented on code in PR #2268:
URL:
https://github.com/apache/incubator-streampark/pull/2268#discussion_r1104623853
##########
streampark-console/streampark-console-service/src/main/java/org/apache/streampark/console/core/task/CheckpointProcessor.java:
##########
@@ -52,30 +62,31 @@ public class CheckpointProcessor {
@Autowired private SavePointService savePointService;
- public void process(Long appId, CheckPoints checkPoints) {
- CheckPoints.Latest latest = checkPoints.getLatest();
- if (latest == null || latest.getCompleted() == null) {
- return;
- }
- CheckPoints.CheckPoint checkPoint = latest.getCompleted();
+ public void process(Long appId, @Nonnull CheckPoints checkPoints) {
+ checkPoints.getLatestCheckpoint().forEach(checkPoint -> process(appId,
checkPoint));
+ }
+
+ private void process(Long appId, @Nonnull CheckPoints.CheckPoint checkPoint)
{
Application application = applicationService.getById(appId);
CheckPointStatus status = checkPoint.getCheckPointStatus();
if (CheckPointStatus.COMPLETED.equals(status)) {
- String cacheId = appId + "_" + application.getJobId();
- Long latestId =
- checkPointCache.get(
- cacheId,
- key -> {
- SavePoint savePoint = savePointService.getLatest(appId);
- return
Optional.ofNullable(savePoint).map(SavePoint::getChkId).orElse(null);
- });
-
- if (latestId == null || latestId < checkPoint.getId()) {
- saveSavepoint(checkPoint, application);
+ String cacheId = getCacheIdForCheckpoint(appId, application.getJobId());
+ Long latestChkId = getLatestCheckpointedId(appId, cacheId);
+ if (shouldStoreAsSavepoint(checkPoint, latestChkId)) {
+ savepointedCache.put(checkPoint.getExternalPath(), DEFAULT_FLAG_BYTE);
Review Comment:
> Why don't using the similar logic to checkpointCache that using the
cacheId as the key and checkpointId as the value?
We can't judge the savepoint if be stored out of TTL. From the point of
view, it doesn't matter what type of value is stored. But the appId + jobId
if shorter than savepoint path as cache key mostly.
I‘m hesitating whether we should take savepoint as more important than
checkpoint. If we treat each other equally, on the contrary, this part of logic
will become simple to share the cache with checkpoint
--
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]