markap14 commented on a change in pull request #3702: NIFI-6636: Fixed
ListGCSBucket file duplication error
URL: https://github.com/apache/nifi/pull/3702#discussion_r321862520
##########
File path:
nifi-nar-bundles/nifi-gcp-bundle/nifi-gcp-processors/src/main/java/org/apache/nifi/processors/gcp/storage/ListGCSBucket.java
##########
@@ -381,40 +384,36 @@ public void onTrigger(ProcessContext context,
ProcessSession session) throws Pro
// Update state
if (lastModified > maxTimestamp) {
maxTimestamp = lastModified;
- currentKeys.clear();
+ maxKeys.clear();
}
if (lastModified == maxTimestamp) {
- currentKeys.add(blob.getName());
+ maxKeys.add(blob.getName());
}
- listCount++;
+ loadCount++;
}
- blobPages = blobPages.getNextPage();
- commit(context, session, listCount);
- listCount = 0;
- } while (blobPages != null);
-
- currentTimestamp = maxTimestamp;
+ commit(context, session, loadCount);
- final long listMillis =
TimeUnit.NANOSECONDS.toMillis(System.nanoTime() - startNanos);
- getLogger().info("Successfully listed GCS bucket {} in {} millis", new
Object[]{bucket, listMillis});
+ blobPage = blobPage.getNextPage();
+ } while (blobPage != null);
- if (!commit(context, session, listCount)) {
- if (currentTimestamp > 0) {
- persistState(context);
- }
- getLogger().debug("No new objects in GCS bucket {} to list.
Yielding.", new Object[]{bucket});
+ if (maxTimestamp != 0) {
+ currentTimestamp = maxTimestamp;
+ currentKeys = maxKeys;
+ persistState(context);
+ } else {
+ getLogger().debug("No new objects in GCS bucket {} to load.
Yielding.", new Object[]{bucket});
context.yield();
}
+
+ final long listMillis =
TimeUnit.NANOSECONDS.toMillis(System.nanoTime() - startNanos);
+ getLogger().info("Successfully listed GCS bucket {} in {} millis", new
Object[]{bucket, listMillis});
}
- private boolean commit(final ProcessContext context, final ProcessSession
session, int listCount) {
- boolean willCommit = listCount > 0;
- if (willCommit) {
- getLogger().info("Successfully listed {} new files from GCS;
routing to success", new Object[] {listCount});
+ private void commit(final ProcessContext context, final ProcessSession
session, int loadCount) {
+ if (loadCount > 0) {
+ getLogger().info("Successfully loaded {} new files from GCS;
routing to success", new Object[] {loadCount});
session.commit();
- persistState(context);
Review comment:
I don't think removing this is a good idea. Consider performing a listing of
a bucket that has 1 million entries. The processor runs and performs a listing
of 800,000 elements, committing the session several times. Now NiFi is
restarted. The state was not persisted. So now, upon restart, NiFi will
duplicate each of those 800,000 elements. The call to `persistState` there
makes sense to avoid huge amounts of duplication when performing a listing on a
huge bucket.
----------------------------------------------------------------
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.
For queries about this service, please contact Infrastructure at:
[email protected]
With regards,
Apache Git Services
