activemq-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Andrew Miller <z2amil...@gmail.com>
Subject Re: IndexOutOfBoundsException in AMQMessageStore using 20071028.173739
Date Fri, 02 Nov 2007 02:16:57 GMT


Thanks for the fix!  We're running the test now...  Parameters:

SVN version: 591163
Client jar: Updated to 5.0-20071028.173739

We haven't seen any of the original IndexOutOfBounds exception messages,
even after about 1.2 million Enqueues.   However, the consumers appear to be
stuck and unable to consume.     We had a RecoveryListenerAdapter message
show up in the log, though I'm not sure this correlates with the inability
of the consumers to consume these messages.

2007-11-01 18:10:42,645 [68.194.21:40095] ERROR RecoveryListenerAdapter       
- Message id ID:rlab-0007.xxxx.com-48174-119396578373
6-0:2:1:1:11 could not be recovered from the data store!

I even wrote a little perl client that does nothing but subscribe to this
queue, consumes one message, and exits.  The script just blocks and then
times out, even though there are many messages in the queue:

rlab-0007:/opt/activemq/bin# ./query -QQueue=IG
ACTIVEMQ_HOME: /opt/activemq
ACTIVEMQ_BASE: /opt/activemq
Type = Queue
DispatchCount = 213
Destination = IG
MaxEnqueueTime = 1371
QueueSize = 926441
Name = IG
DequeueCount = 213
MemoryPercentageUsed = 0
ConsumerCount = 19
MemoryLimit = 30198988
EnqueueCount = 1190424
MinEnqueueTime = 1
AverageEnqueueTime = 196.80751173708921

Here's the one-off perl script and the output: (I assume there's no reason I
would be unable to consume a message sent via JMS in stomp/perl?) 
(redacting script since my previous mail on this subject got rejected for
being too spammy - suffice to say it simply subscribes to the queue in
question and does a stomp->receive_frame())

amiller@mirror-01:/opt/amiller$ ./c1.pl 
connected 
could not receive message after 20 seconds at ./c1.pl line 17, <GEN0> line
4.

We also had this problem with the 20071031 snapshot, but at that time I
thought it was because we were using the 4.1.1 client jars.  BTW even though
we didn't see this issue in 20071002, we needed to upgrade to solve
AMQ-1479.

Thanks again!
-Andy




rajdavies wrote:
> 
> There's a potential fix for this I've just added to trunk - would it  
> be possible to build the latest and test ?
> 
> 
> cheers,
> 
> Rob
> 
> http://rajdavies.blogspot.com/
> 
> 
> 
> On Nov 1, 2007, at 9:47 PM, Andrew Miller wrote:
> 
>>
>> We've been using ActiveMQ in an infrastructure refresh project, and  
>> have run
>> into a nasty looking bug, it looks like we are dropping  
>> messages.    We have
>> a few fast producers that will produce approximately 4-5 million  
>> persistent
>> messages over the course of about 2.5 hours, while many slower  
>> consumers
>> will consume these messages over the course of a day or so.
>>
>> We are using Jencks/Spring for the producers and consumers, using the
>> openwire protocol.  For this test, we were using the AMQ 4.1.1  
>> client jars,
>> if that matters.  (We will try updating to the new AMQ5 client jars  
>> today)
>>
>> About halfway through the production run, we started getting
>> IndexOutOfBounds exceptions in the log pretty frequently.  Later on  
>> the
>> consumer tries to get these messages but appears to fail.
> 
> (exception messages trimmed)
> 
> 

-- 
View this message in context: http://www.nabble.com/IndexOutOfBoundsException-in-AMQMessageStore-using-20071028.173739-tf4734264s2354.html#a13541167
Sent from the ActiveMQ - Dev mailing list archive at Nabble.com.


Mime
View raw message