saintstack commented on a change in pull request #974: HBASE-23587 The
FSYNC_WAL flag does not work on branch-2.x
URL: https://github.com/apache/hbase/pull/974#discussion_r362637288
##########
File path:
hbase-server/src/main/java/org/apache/hadoop/hbase/regionserver/wal/FSHLog.java
##########
@@ -577,10 +577,9 @@ public void run() {
//TraceScope scope = Trace.continueSpan(takeSyncFuture.getSpan());
long start = System.nanoTime();
Throwable lastException = null;
- boolean wasRollRequested = false;
try {
TraceUtil.addTimelineAnnotation("syncing writer");
- writer.sync(useHsync);
+ writer.sync(takeSyncFuture.isForceSync());
Review comment:
Ok. I went back over this and this is correct as the SyncFuture uses
whatever the config is set to when created.
bq. So the desion allow different table/mutation use different durability,
but these table share one WAL writer?
Yes.. this could make for some 'interesting' results. IIRC, looking at this
w/ Lars a while back, if async WAL and you were trying to manage the syncs from
the application, a client could get an ack though all puts hadn't yet syncd'.
----------------------------------------------------------------
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
