DistributedUpdateProcessor.java
@Override
public void finish() throws IOException {
assert ! finished : "lifecycle sanity check";
finished = true;
if (zkEnabled) doFinish();
if (next != null && nodes == null) next.finish();
}
HdfsTranscationLog.finish() will call hflush or hsync to save tlog.
But when we have two replicas, the value of nodes is not equal to null. So,
next.finish() will not be executed.
--
View this message in context:
http://lucene.472066.n3.nabble.com/why-leader-replica-does-not-call-HdfsTransactionLog-finish-tp4327139.html
Sent from the Solr - User mailing list archive at Nabble.com.
