update - it has been fixed in 0.98.4, HBASE-11374
On Sat, Aug 9, 2014 at 5:53 PM, Qiang Tian <[email protected]> wrote: > please see HBASE-11714 > > > On Sat, Aug 9, 2014 at 4:27 PM, Qiang Tian <[email protected]> wrote: > >> took a quick look. it looks a bug, will dig more and update later. >> >> >> >> On Sat, Aug 9, 2014 at 1:45 AM, Walter King <[email protected]> wrote: >> >>> awesome, thanks! >>> the resets and such were actually what I've been trying to investigate. >>> Recently switched from 0.94 and 0.98, and finding that periodically >>> things >>> are having issues - lots of retry exceptions. Don't have much >>> information, >>> still looking into it when saw this. >>> >>> By timestamp, this seems to be the corresponding client exception: >>> >>> 2014-08-08 17:22:43 o.a.h.h.c.AsyncProcess [INFO] #105158, >>> table=rt_global_monthly_campaign_deliveries, attempt=10/35 failed 500 >>> ops, >>> last exception: java.net.SocketTimeoutException: Call to >>> ip-10-201-128-23.us-west-1.compute.internal/10.201.128.23:60020 failed >>> because java.net.SocketTimeoutException: 2000 millis timeout while >>> waiting >>> for channel to be ready for read. ch : >>> java.nio.channels.SocketChannel[connected local=/10.248.130.152:46014 >>> remote=ip-10-201-128-23.us-west-1.compute.internal/10.201.128.23:60020] >>> on >>> ip-10-201-128-23.us-west-1.compute.internal,60020,1405642103651, tracking >>> started Fri Aug 08 17:21:55 UTC 2014, retrying after 10043 ms, replay 500 >>> ops. >>> >>> 2014-08-08 17:22:53 o.a.h.h.c.AsyncProcess [INFO] #105158, waiting for >>> some >>> tasks to finish. Expected max=0, tasksSent=11, tasksDone=10, >>> currentTasksDone=10, retries=10 hasError=false, >>> tableName=rt_global_monthly_campaign_deliveries >>> >>> >>> On Fri, Aug 8, 2014 at 2:15 AM, Esteban Gutierrez <[email protected]> >>> wrote: >>> >>> > The RequestContext seems to be ok since it hasn't been set. So just >>> > decrementing the call size should be enough. >>> > >>> > cheers, >>> > esteban. >>> > >>> > >>> > -- >>> > Cloudera, Inc. >>> > >>> > >>> > >>> > On Fri, Aug 8, 2014 at 2:10 AM, Esteban Gutierrez < >>> [email protected]> >>> > wrote: >>> > >>> > > >>> > > Looks like after HBASE-10506 we didn't do what CallRunner used to do >>> > > before: clear the RequestContext and decrement callQueueSize. >>> > > >>> > > Created HBASE-11705 to address this issue. >>> > > >>> > > >>> > > >>> > > >>> > > >>> > > >>> > > -- >>> > > Cloudera, Inc. >>> > > >>> > > >>> > > >>> > > On Fri, Aug 8, 2014 at 1:56 AM, Qiang Tian <[email protected]> >>> wrote: >>> > > >>> > >> good point. that is a big suspect. >>> > >> >>> > >> I check your log, ClosedChannelException should be triggered by >>> > >> call.sendResponseIfReady()(it is the only request in the queue, so >>> > >> handler >>> > >> send response directly), but at that point the callqueueSize has >>> been >>> > >> decremented. >>> > >> >>> > >> 2014-08-05 00:50:06,727 WARN [RpcServer.handler=57,port=60020] >>> > >> ipc.RpcServer (RpcServer.java:processResponse(1041)) - >>> > >> RpcServer.respondercallId: 118504 service: ClientService methodName: >>> > Multi >>> > >> size: 141.9 K connection: 10.248.134.67:55347: output error >>> > >> 2014-08-05 00:50:06,727 WARN [RpcServer.handler=57,port=60020] >>> > >> ipc.RpcServer (CallRunner.java:run(135)) - >>> > >> RpcServer.handler=57,port=60020: >>> > >> caught a ClosedChannelException, this means that the server was >>> > processing >>> > >> a request but the client went away. The error message was: null >>> > >> >>> > >> it looks you have got the fix, would you file a jira? >>> > >> thanks. >>> > >> >>> > >> >>> > >> On Fri, Aug 8, 2014 at 2:41 PM, Walter King <[email protected]> >>> wrote: >>> > >> >>> > >> > I've only looked at the code a little, and likely missed >>> something, >>> > but >>> > >> > does this if block decrement the call queue, if the client already >>> > >> closed >>> > >> > the connection? >>> > >> > >>> > >> > >>> > >> > >>> > >> >>> > >>> https://github.com/apache/hbase/blob/07a771866f18e8ec532c14f624fa908815bd88c7/hbase-server/src/main/java/org/apache/hadoop/hbase/ipc/CallRunner.java#L74 >>> > >> > >>> > >> > >>> > >> > >>> > >> > On Thu, Aug 7, 2014 at 11:32 PM, Walter King <[email protected]> >>> > wrote: >>> > >> > >>> > >> > > Yes, sorry, CallQueueTooBigException. but that value never >>> returns >>> > to >>> > >> > > zero, even when number of requests goes to zero. The call >>> queue too >>> > >> big >>> > >> > > happens if any regionserver is up for a long enough period of >>> time, >>> > >> so I >>> > >> > > have to periodically restart them. Also at that 15:30 time I >>> wasn't >>> > >> > > seeing that exception, but it seems like that is one time in >>> which a >>> > >> call >>> > >> > > didnt properly decrement the callqueuesize because it was at >>> zero >>> > >> before >>> > >> > > and has never hit zero again - today the minimum is even higher. >>> > >> > > >>> > >> > > >>> > >> > > On Thu, Aug 7, 2014 at 9:14 PM, Qiang Tian <[email protected]> >>> > wrote: >>> > >> > > >>> > >> > >> bq. "Eventually we ran into ipc queue size full messages being >>> > >> returned >>> > >> > to >>> > >> > >> clients trying large batch puts, as it approaches a gigabyte." >>> > >> > >> >>> > >> > >> Do you mean CallQueueTooBigException? it looks not the queue >>> size, >>> > >> but >>> > >> > the >>> > >> > >> data size that client sends..configured by >>> > >> > >> "hbase.ipc.server.max.callqueue.size". >>> > >> > >> >>> > >> > >> I guess when you client got the exception, it closed the >>> exception >>> > >> and >>> > >> > >> causing other shared connection RPC failed. >>> > >> > >> >>> > >> > >> >>> > >> > >> 2014-08-06 22:27:57,253 WARN [RpcServer.reader=9,port=60020] >>> > >> > >> ipc.RpcServer >>> > >> > >> (RpcServer.java:doRead(794)) - RpcServer.listener,port=60020: >>> count >>> > >> of >>> > >> > >> bytes read: 0 >>> > >> > >> java.io.IOException: Connection reset by peer >>> > >> > >> at sun.nio.ch.FileDispatcherImpl.read0(Native Method) >>> > >> > >> at sun.nio.ch.SocketDispatcher.read(SocketDispatcher.java:39) >>> > >> > >> at sun.nio.ch.IOUtil.readIntoNativeBuffer(IOUtil.java:223) >>> > >> > >> at sun.nio.ch.IOUtil.read(IOUtil.java:197) >>> > >> > >> at >>> sun.nio.ch.SocketChannelImpl.read(SocketChannelImpl.java:379) >>> > >> > >> at >>> > >> > >>> org.apache.hadoop.hbase.ipc.RpcServer.channelRead(RpcServer.java:2229) >>> > >> > >> at >>> > >> > >> >>> > >> > >> >>> > >> > >>> > >> >>> > >>> org.apache.hadoop.hbase.ipc.RpcServer$Connection.readAndProcess(RpcServer.java:1415) >>> > >> > >> at >>> > >> > >> >>> > >> > >>> > >> >>> > >>> org.apache.hadoop.hbase.ipc.RpcServer$Listener.doRead(RpcServer.java:790) >>> > >> > >> at >>> > >> > >> >>> > >> > >> >>> > >> > >>> > >> >>> > >>> org.apache.hadoop.hbase.ipc.RpcServer$Listener$Reader.doRunLoop(RpcServer.java:581) >>> > >> > >> at >>> > >> > >> >>> > >> > >> >>> > >> > >>> > >> >>> > >>> org.apache.hadoop.hbase.ipc.RpcServer$Listener$Reader.run(RpcServer.java:556) >>> > >> > >> at >>> > >> > >> >>> > >> > >> >>> > >> > >>> > >> >>> > >>> java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1145) >>> > >> > >> at >>> > >> > >> >>> > >> > >> >>> > >> > >>> > >> >>> > >>> java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:615) >>> > >> > >> at java.lang.Thread.run(Thread.java:744) >>> > >> > >> 2014-08-06 22:27:57,257 WARN [RpcServer.handler=18,port=60020] >>> > >> > >> ipc.RpcServer (RpcServer.java:processResponse(1041)) - >>> > >> > >> RpcServer.respondercallId: 84968 service: ClientService >>> methodName: >>> > >> > Multi >>> > >> > >> size: 17.7 K connection: 10.248.130.152:49780: output error >>> > >> > >> 2014-08-06 22:27:57,258 WARN [RpcServer.handler=18,port=60020] >>> > >> > >> ipc.RpcServer (CallRunner.java:run(135)) - >>> > >> > >> RpcServer.handler=18,port=60020: >>> > >> > >> caught a ClosedChannelException, this means that the server was >>> > >> > processing >>> > >> > >> a request but the client went away. The error message was: null >>> > >> > >> 2014-08-06 22:27:57,260 WARN [RpcServer.handler=61,port=60020] >>> > >> > >> ipc.RpcServer (RpcServer.java:processResponse(1041)) - >>> > >> > >> RpcServer.respondercallId: 83907 service: ClientService >>> methodName: >>> > >> > Multi >>> > >> > >> size: 17.1 K connection: 10.248.1.56:53615: output error >>> > >> > >> 2014-08-06 22:27:57,263 WARN [RpcServer.handler=61,port=60020] >>> > >> > >> ipc.RpcServer (CallRunner.java:run(135)) - >>> > >> > >> RpcServer.handler=61,port=60020: >>> > >> > >> caught a ClosedChannelException, this means that the server was >>> > >> > processing >>> > >> > >> a request but the client went away. The error message was: null >>> > >> > >> >>> > >> > >> >>> > >> > >> >>> > >> > >> On Fri, Aug 8, 2014 at 2:57 AM, Walter King <[email protected] >>> > >>> > >> wrote: >>> > >> > >> >>> > >> > >> > >>> > >> > >> >>> > >> > >>> > >> >>> > >>> https://gist.github.com/walterking/4c5c6f5e5e4a4946a656#file-gistfile1-txt >>> > >> > >> > >>> > >> > >> > >>> > >> > >>> > http://adroll-test-sandbox.s3.amazonaws.com/regionserver.stdout.log.gz >>> > >> > >> > >>> > >> > >> > These are logs from that particular server, and the debug >>> dump >>> > from >>> > >> > >> now(no >>> > >> > >> > restart in between). The times in the graph are pacific, so >>> it >>> > >> should >>> > >> > >> be >>> > >> > >> > around 2014-08-06 22:25:00. I do see some exceptions around >>> > there. >>> > >> > >> > >>> > >> > >> >>> > >> > > >>> > >> > > >>> > >> > >>> > >> >>> > > >>> > > >>> > >>> >> >> >
