flume-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Brock Noland <br...@cloudera.com>
Subject Re: Custom sink - "close() called when transaction is OPEN" error
Date Fri, 16 Nov 2012 15:32:00 GMT
Would you be able to send the source of your sink via pastbin in
addition to your config?

On Fri, Nov 16, 2012 at 9:21 AM, Andrew Jones <andrewjones86@gmail.com> wrote:
> I tried logging the first throwable, but now that is just the
> IllegalStateException.
>
> Today I have been looking at Flume-1.3.0rc3 and I have noticed the same
> problem. This is using the Avro source, File channel and our custom sink.
> After Flume reloads its config, the first error message comes when the Avro
> source starts up:
>
> 16 Nov 2012 16:04:25,237 INFO  [lifecycleSupervisor-1-4]
> (org.apache.flume.source.AvroSource.start:142)  - Starting Avro source
> source: { bindAddress: 0.0.0.0, port: 36060 }...
> 16 Nov 2012 16:04:25,484 ERROR
> [SinkRunner-PollingRunner-DefaultSinkProcessor]
> (org.apache.flume.SinkRunner$PollingRunner.run:160)  - Unable to deliver
> event. Exception follows.
> java.lang.IllegalStateException: close() called when transaction is OPEN -
> you must either commit or rollback first
>         at
> com.google.common.base.Preconditions.checkState(Preconditions.java:176)
>         at
> org.apache.flume.channel.BasicTransactionSemantics.close(BasicTransactionSemantics.java:179)
>         at
> com.arm.pd.brodie.flume.sink.ResultSink.processSingle(ResultSink.java:440)
>         at
> com.arm.pd.brodie.flume.sink.ResultSink.process(ResultSink.java:172)
>         at
> org.apache.flume.sink.DefaultSinkProcessor.process(DefaultSinkProcessor.java:68)
>         at
> org.apache.flume.SinkRunner$PollingRunner.run(SinkRunner.java:147)
>         at java.lang.Thread.run(Thread.java:636)
> 16 Nov 2012 16:04:25,594 INFO  [lifecycleSupervisor-1-4]
> (org.apache.flume.instrumentation.MonitoredCounterGroup.register:89)  -
> Monitoried counter group for type: SOURCE, name: source, registered
> successfully.
> 16 Nov 2012 16:04:25,595 INFO  [lifecycleSupervisor-1-4]
> (org.apache.flume.instrumentation.MonitoredCounterGroup.start:73)  -
> Component type: SOURCE, name: source started
> 16 Nov 2012 16:04:25,595 INFO  [lifecycleSupervisor-1-4]
> (org.apache.flume.source.AvroSource.start:168)  - Avro source source
> started.
>
> I then continually get errors from the Sink, presumably as its been called
> periodically to check for events in the channel. So is it possible its the
> Avro source causing the issue?
>
> There should have been nothing persisted in the file channel when
> restarting.
>
> When the transaction gets messed up like this, is there a way to refresh it,
> preferably without losing any data?
>
> I am still able to send things to flume and they get processed and inserted
> by my sink, so it still seems to work OK.
>
> Thanks,
> Andrew
>
>
>
> On 15 November 2012 12:50, Brock Noland <brock@cloudera.com> wrote:
>>
>> Can you log the Throwable as the first thing in the catch block to see
>> if something and what it is, is being thrown?
>>
>> Transactions are thread local so if for some reason the the sequencing
>> gets messed up on an earlier call the process, every call to
>> transaction will thrown an exception including begin.
>>
>>
>> https://github.com/apache/flume/blob/trunk/flume-ng-core/src/main/java/org/apache/flume/channel/BasicTransactionSemantics.java
>>
>> As I stated in FLUME-1089 I think that when close is called it should
>> forcefully destroy the transaction like JDBC close() but I have not
>> got much agreement.
>>
>>
>> On Thu, Nov 15, 2012 at 5:24 AM, Andrew Jones <andrewjones86@gmail.com>
>> wrote:
>> > We are using Flume 1.2.0. We have a custom source, although it passes
>> > through an Avro Sink and Source before getting to the sink. We are now
>> > using
>> > the memory channel, although had just switched from the JDBC channel
>> > when we
>> > started seeing these errors, so maybe that's something to do with it?
>> >
>> > I tried wrapping transaction.rollback(); in a try catch and logging in
>> > the
>> > catch, but it wasn't called, so I don't think the rollback is throwing
>> > an
>> > error.
>> >
>> > I think it may have something to do with switching channels, as right
>> > after
>> > Flume reloaded the config we started getting errors. I have restarted
>> > the
>> > flume node manually and we are still getting the error.
>> >
>> > Thanks,
>> > Andrew
>> >
>> >
>> > On 14 November 2012 20:02, Hari Shreedharan <hshreedharan@cloudera.com>
>> > wrote:
>> >>
>> >> Which version of Flume are you using? It looks like the transaction was
>> >> never rolled back or committed. It is likely that the rollback method
>> >> too
>> >> threw some exception, and the rollback was not successful. Also, what
>> >> channel are you using?
>> >>
>> >>
>> >> Thanks,
>> >> Hari
>> >>
>> >> --
>> >> Hari Shreedharan
>> >>
>> >> On Wednesday, November 14, 2012 at 8:55 AM, Andrew Jones wrote:
>> >>
>> >> Hi,
>> >>
>> >> I have a custom sink which has been working fine, but recently I have
>> >> started seeing this error in the logs:
>> >>
>> >> Unable to deliver event. Exception follows.
>> >> java.lang.IllegalStateException: close() called when transaction is
>> >> OPEN -
>> >> you must either commit or rollback first
>> >>         at
>> >> com.google.common.base.Preconditions.checkState(Preconditions.java:176)
>> >> ...
>> >>
>> >>
>> >> After having a google and finding
>> >> https://issues.apache.org/jira/browse/FLUME-1089, I have double checked
>> >> I am
>> >> using the correct try, catch, finally idiom that other sinks use, and I
>> >> seem
>> >> to be doing the same. I do the following:
>> >>
>> >> public Status process() throws EventDeliveryException {
>> >> Status status = Status.READY;
>> >>
>> >> Channel channel = getChannel();
>> >> Transaction transaction = channel.getTransaction();
>> >>
>> >> try {
>> >> transaction.begin();
>> >>
>> >>                         // does a bit of processing and
>> >>                         // writes out the event to MongoDB
>> >>
>> >>                         transaction.commit();
>> >>
>> >> } catch (Throwable t) {
>> >> transaction.rollback();
>> >>
>> >> if (t instanceof Error) {
>> >> throw (Error) t;
>> >> } else if  (t instanceof EventDeliveryException) {
>> >> throw (EventDeliveryException) t;
>> >> } else if (t instanceof ChannelException) {
>> >> logger.error("Brodie Log Sink " + getName() + ": Unable to get event
>> >> from"
>> >> +
>> >> " channel " + channel.getName() + ". Exception follows.", t);
>> >> status = Status.BACKOFF;
>> >> } else {
>> >> throw new EventDeliveryException("Failed to send events", t);
>> >> }
>> >> } finally {
>> >> transaction.close();
>> >> }
>> >>
>> >> return status;
>> >> }
>> >>
>> >> }
>> >>
>> >> All of this code came from looking at other sinks (Avro and HDFS), so I
>> >> am
>> >> pretty sure its correct.
>> >>
>> >> Can anyone see anything that might be a problem, or is there anything
>> >> else
>> >> I can do to avoid this error?
>> >>
>> >> Thanks,
>> >> Andrew
>> >>
>> >>
>> >
>>
>>
>>
>> --
>> Apache MRUnit - Unit testing MapReduce -
>> http://incubator.apache.org/mrunit/
>
>



-- 
Apache MRUnit - Unit testing MapReduce - http://incubator.apache.org/mrunit/

Mime
View raw message