dlmarion commented on a change in pull request #2132:
URL: https://github.com/apache/accumulo/pull/2132#discussion_r644897548
##########
File path:
server/compaction-coordinator/src/main/java/org/apache/accumulo/coordinator/DeadCompactionDetector.java
##########
@@ -85,31 +103,55 @@ private void detectDeadCompactions() {
running.forEach((ecid) -> {
if (tabletCompactions.remove(ecid) != null) {
- log.trace("Removed {} running on a compactor", ecid);
+ log.trace("Removed compaction {} running on a compactor", ecid);
+ }
+ if (this.deadCompactions.remove(ecid) != null) {
+ log.trace("Removed {} from the dead compaction map, it's running on a
compactor", ecid);
}
});
// Determine which compactions are currently committing and remove those
context.getAmple().getExternalCompactionFinalStates()
- .map(ecfs ->
ecfs.getExternalCompactionId()).forEach(tabletCompactions::remove);
+ .map(ecfs -> ecfs.getExternalCompactionId()).forEach(ecid -> {
+ if (tabletCompactions.remove(ecid) != null) {
+ log.trace("Removed compaction {} that is committing", ecid);
+ }
+ if (this.deadCompactions.remove(ecid) != null) {
+ log.trace("Removed {} from the dead compaction map, it's
committing", ecid);
+ }
+ });
- tabletCompactions
- .forEach((ecid, extent) -> log.debug("Detected dead compaction {} {}",
ecid, extent));
+ tabletCompactions.forEach((ecid, extent) -> {
+ log.debug("Possible dead compaction detected {} {}", ecid, extent);
+ this.deadCompactions.putIfAbsent(ecid, System.currentTimeMillis());
+ });
// Everything left in tabletCompactions is no longer running anywhere and
should be failed.
// Its possible that a compaction committed while going through the steps
above, if so then
// that is ok and marking it failed will end up being a no-op.
- try {
- coordinator.compactionFailed(tabletCompactions);
- } catch (UnknownCompactionIdException e) {
- // One or more Ids was not in the Running compaction list. This is ok to
ignore.
- }
+ long now = System.currentTimeMillis();
+ this.deadCompactions.forEach((eci, startTime) -> {
+ if ((now - startTime) > threshold) {
+ // Compaction believed to be dead for two cycles. Fail it.
+ try {
+ log.warn(
+ "Failing compaction {} which is believed to be dead. Last seen
at {} and not seen since.",
+ eci, startTime);
+ coordinator.compactionFailed(tabletCompactions);
+ this.deadCompactions.remove(eci);
+ } catch (UnknownCompactionIdException e) {
+ // One or more Ids was not in the Running compaction list. This is
ok to ignore.
+ }
+ }
+ });
Review comment:
I applied this change in 95e8fec149722ac3f205734103223e3f42b33438
##########
File path:
server/compaction-coordinator/src/main/java/org/apache/accumulo/coordinator/DeadCompactionDetector.java
##########
@@ -85,31 +103,55 @@ private void detectDeadCompactions() {
running.forEach((ecid) -> {
if (tabletCompactions.remove(ecid) != null) {
- log.trace("Removed {} running on a compactor", ecid);
+ log.trace("Removed compaction {} running on a compactor", ecid);
+ }
+ if (this.deadCompactions.remove(ecid) != null) {
+ log.trace("Removed {} from the dead compaction map, it's running on a
compactor", ecid);
}
});
// Determine which compactions are currently committing and remove those
context.getAmple().getExternalCompactionFinalStates()
- .map(ecfs ->
ecfs.getExternalCompactionId()).forEach(tabletCompactions::remove);
+ .map(ecfs -> ecfs.getExternalCompactionId()).forEach(ecid -> {
+ if (tabletCompactions.remove(ecid) != null) {
+ log.trace("Removed compaction {} that is committing", ecid);
+ }
+ if (this.deadCompactions.remove(ecid) != null) {
+ log.trace("Removed {} from the dead compaction map, it's
committing", ecid);
+ }
+ });
- tabletCompactions
- .forEach((ecid, extent) -> log.debug("Detected dead compaction {} {}",
ecid, extent));
+ tabletCompactions.forEach((ecid, extent) -> {
+ log.debug("Possible dead compaction detected {} {}", ecid, extent);
+ this.deadCompactions.putIfAbsent(ecid, System.currentTimeMillis());
Review comment:
I applied this change in 95e8fec149722ac3f205734103223e3f42b33438
--
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]
