> My understanding is that HBase requires "durable sync" capabilities of HDFS (i.e. hflush() hsync()), but does *not* require file append capabilities.
99.99% true. The remaining 0.01% is an exceptional code path during the data recovery (as a fall back mechanism to ensure that we can start the recovery and nobody is writing). This code was very unlikely to be called. It was removed in august this year (in HBASE-8204), for both 0.94 & 0.96. > In an HDFS 2.x environment, will HBase function properly with 'dfs.support.append' set to FALSE? Yes, you don't need it. Cheers, Nicolas On Thu, Oct 24, 2013 at 4:01 PM, Sirianni, Eric <eric.siria...@netapp.com>wrote: > I am aware of the history of HBase and HDFS "append" and have read the > various blog posts and JIRAs. In particular: > https://issues.apache.org/jira/browse/HBASE-5676 > https://issues.apache.org/jira/browse/HDFS-3120 > > My understanding is that HBase requires "durable sync" capabilities of > HDFS (i.e. hflush() hsync()), but does *not* require file append > capabilities. > > I'm looking for a quick confirmation of my understanding *in a Hadoop 2.x > context*. > > In an HDFS 2.x environment, will HBase function properly with > 'dfs.support.append' set to FALSE? > > Eric >
