jvrao commented on a change in pull request #853: BP-14 WriteFlag DEFERRED_SYNC
Client Side Implementation
URL: https://github.com/apache/bookkeeper/pull/853#discussion_r188504462
##########
File path:
bookkeeper-server/src/main/java/org/apache/bookkeeper/client/LedgerHandle.java
##########
@@ -1678,17 +1692,23 @@ void sendAddSuccessCallbacks() {
return;
}
// Check if it is the next entry in the sequence.
- if (pendingAddOp.entryId != 0 && pendingAddOp.entryId !=
lastAddConfirmed + 1) {
+ if (pendingAddOp.entryId != 0 && pendingAddOp.entryId !=
pendingAddsSequenceHead) {
if (LOG.isDebugEnabled()) {
- LOG.debug("Head of the queue entryId: {} is not lac: {} +
1", pendingAddOp.entryId,
- lastAddConfirmed);
+ LOG.debug("Head of the queue entryId: {} is not the
expected value: {}", pendingAddOp.entryId,
+ pendingAddsSequenceHead);
}
return;
}
pendingAddOps.remove();
explicitLacFlushPolicy.updatePiggyBackedLac(lastAddConfirmed);
- lastAddConfirmed = pendingAddOp.entryId;
+ if (writeFlags.contains(WriteFlag.DEFERRED_SYNC)) {
+ // cannot update LastAddConfirmed, we are not sure bookie has
stored the entry durably at this point
+ } else {
+ this.lastAddConfirmed = Math.max(lastAddConfirmed,
pendingAddOp.entryId);
+ }
Review comment:
Why are you changing this logic? Do you ever see a situation where
lastAddConfirmed > pendingAddOp.entryId? If so then there is a problem right
there. I would assert here rather than get Math.max. Also why can't you just
check for !DEFERRED_SYNC?
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on 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
