1) You probably want to upgrade to a more recent version of HBase 2) You probably want to read this: http://hbase.apache.org/book.html#performance
Not knowing anything about your cluster size or table design, Put delays can be exacerbated by a number of things. 3) As for the time limit, I think you'd be better off dropping the writes off in a queue (i.e., fire-n-forget) and letting HBase do its job than trying to kill a Put because it went over a time limit and then immediately re-trying. But that's an off-the-cuff opinion. -----Original Message----- From: Douglas Campbell [mailto:deegs...@yahoo.com] Sent: Thursday, June 02, 2011 5:09 PM To: hbase users Subject: Timeouts on gets and puts Is there an easy way to control this? some magic setting in the HBAseConfiguration passed to HTable? The problem we're facing is that for each put and get we block for however long it takes to do the put. The norm put/get is fast or at least impressive to me and yet some puts take 10sec + We're running 20.6. The plan is to impose a time limit on each read write op, retry if limit is hit x amount of times. In this way we know that given an input set of size N, it will complete, possibly with failures, in a bounded amount of time.