hbase-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Patrick Schless <patrick.schl...@gmail.com>
Subject ageOfLastAppliedOp, ageOfLastShippedOp
Date Fri, 23 Aug 2013 19:14:36 GMT
We run two hbase clusters, and one (master) replicates to the other
(standby). We did some maintenance last night which involved bringing all
of hbase down while we made changes to HDFS. After bringing things back up,
our ageOfLastShippedOp on a few of the master region servers jumped to
around -9 petaseconds, and the ageOfLastShippedOp on one of the slave
region servers jumped to a little over 800 seconds.

We fixed the -9 petaseconds issue (which I assume was a reporting problem)
by restarting the handful of (master) region servers that had that issue.

For the slave region server with 800 seconds of ageOfLastAppliedOp, I've
restarted the process, but the value didn't change. I assume that nobody in
the master cluster is replicating to this node, so the value will just stay
at 800. This is a problem, since we have alerts that trigger on high
values. Is there any way to address this?


We're on hbase 0.92 (CHD4.1)

  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message