Thanks Josh/Keith. I doubt GC is the culprit here, I suspect something wrong with the Accumulo client connection. I set the JVM size to 6GB, and right now it's only using 2.5GB I just sent request from client, and the time it took to process the request and call Accumulo (and get back the read results) took more than 8s, when the same request going to another instance, that whole thing only took ~27ms for example (see how big the diff here).
If there is no way to inspect this from Accumulo side, then I plan to kill that instance, and see if the same issue may happen again in the future. This is first time I experienced this (it could be one off-issue), which I can't reproduce, so not related to connector caching or not. Thanks, Z -- View this message in context: http://apache-accumulo.1065345.n5.nabble.com/cached-connector-tp16202p16219.html Sent from the Developers mailing list archive at Nabble.com.
