You may be running into this issue: https://issues.apache.org/jira/browse/ACCUMULO-1572
However, it is normally unusual to get a disconnected event from zookeeper. Could you describe your environment a bit? Please provide versions of hadoop, zookeeper, accumulo, OS and native/virtual execution. -Eric On Mon, Nov 4, 2013 at 2:24 PM, buttercream <[email protected]>wrote: > I'm having a consistent problem with my tservers being killed by the > master. > One by one they turn off over the course of a week or so. In the tserver > log, I see a debug message right before the shutdown that says: > Unexpected event watching lock node WatchedEvent state:Disconnected > type:None path: null > > Looking at the timing of what led up to this, the tserver has the following > events: > 02:03:37,922 DEBUG: gc > 02:03:38,341 DEBUG: UpSess <ipaddress> > 02:03:38,585 DEBUG: MultiScanSess > 02:03:38,596 DEBUG: MultiScanSess > 02:03:38,601 DEBUG: MultiScanSess > 02:04:14,007 DEBUG: Unexpected event > > In the master log, I see the KeeperException$ConnectionLossException to the > machine that occurs at 02:04:14,017 > > So, is this indicative of just a timeout and if so, is there a setting in > the master to increase the timeout? > > > > -- > View this message in context: > http://apache-accumulo.1065345.n5.nabble.com/Tserver-kills-themselves-from-lost-Zookeeper-locks-tp6125.html > Sent from the Users mailing list archive at Nabble.com. >
