veghlaci05 commented on a change in pull request #3034:
URL: https://github.com/apache/hive/pull/3034#discussion_r820470355
##########
File path: ql/src/java/org/apache/hadoop/hive/ql/txn/compactor/Cleaner.java
##########
@@ -288,14 +285,30 @@ private void clean(CompactionInfo ci, long minOpenTxnGLB,
boolean metricsEnabled
if (metricsEnabled) {
Metrics.getOrCreateCounter(MetricsConstants.COMPACTION_CLEANER_FAILURE_COUNTER).inc();
}
- txnHandler.markFailed(ci);
- } finally {
+ handleCleanerAttemptFailure(ci);
+ } finally {
if (metricsEnabled) {
perfLogger.perfLogEnd(CLASS_NAME, cleanerMetric);
}
}
}
+ private void handleCleanerAttemptFailure(CompactionInfo ci) throws
MetaException {
+ long defaultRetention = getTimeVar(conf,
HIVE_COMPACTOR_CLEANER_RETRY_RETENTION_TIME, TimeUnit.MILLISECONDS);
+ int cleanAttempts = 0;
+ if (ci.retryRetention > 0) {
+ cleanAttempts = (int)(Math.log(ci.retryRetention / defaultRetention) /
Math.log(2)) + 1;
+ }
+ if (cleanAttempts >= getIntVar(conf,
HIVE_COMPACTOR_CLEANER_MAX_RETRY_ATTEMPTS)) {
+ //Mark it as failed if the max attempt threshold is reached.
+ txnHandler.markFailed(ci);
+ } else {
+ //Calculate retry retention time and update record.
+ ci.retryRetention = (long)Math.pow(2, cleanAttempts) * defaultRetention;
Review comment:
Yes, the plan was to have exponential backoff, the wait time doubles in
every iteration (`2^n * 5m` where `n` is the number of attempts). Originally I
wanted to add the backoff value to the last retry attempt time, but that would
need one more new field. So to keep it simple I add it to the submit time.
--
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]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
