On Fri, Feb 26, 2016 at 3:27 PM, z11373 <[email protected]> wrote: > I want to add another point from the observation I just did. > The code in my service will cache the connector based on specific id (let > call it firm id in this case), so what I just did was to send another > request with different firm id to that specific service instance. I can see > from my application log, it creates new connector (and then cache it). > Then I keep sending request with that second firm id, and got expected > processing time. > Then I tried sending request with the first firm id (the one which is > already cached and in 'bad' state), and it consistently taking long time > and > returns after 8s. >
For the slow case, would be nice if you send the request and then capture a few jstacks of the client process. I would like to see those if you can share. Accumulo client code has a good bit of trace level logging. If you can change the log4j config while that client processing is running and then submit the slow request, you may see something useful. > > This test result kind of telling me that the connection for that particular > connector is likely already busted (somehow), but it works fine on new > connector. > > > Thanks, > Z > > > > -- > View this message in context: > http://apache-accumulo.1065345.n5.nabble.com/cached-connector-tp16202p16230.html > Sent from the Developers mailing list archive at Nabble.com. >
