Problem with setting memoryUsage when using persistent messaging with slow consumers

[Duro <du...@zoznam.sk>]

Hello,
i have some problems with configuration when using persistent queue. I have
following scenario:
- broker running on SUSE Linux Enterprise Server 10 SP1 (x86_64)
- Xmx for broker VM set to 800MB
- tcp transport
- client messages (over 500k msgs * 20kB per msgs)
- 4 receivers with Client_Ack wich recovers session after every message (I
need this when my endpoint - http receiver is down),5 delivery attempts,
1000 ms retry timeout, prefetch size 10,
- consuming messages from DLQ immediately (after redelivery)

I've tried 3 different broker configurations with different results:

1. <memoryUsage limit="403mb"/>
result: 
- heap space usage quickly reaches configured maximum and jConsole shows in
MemoryPercentUsage on queue 70
- Warn message and NPE in activemq-startup.log :
WARN  RegionBroker                   - Caught an exception sending to DLQ:
Message ID:txdsa2-b-42053-1242741582775-0:10:5:1:1 dropped=false locked=true
http://www.nabble.com/file/p23657227/403amq.log 403amq.log 
-negtive size of DLQ (-2)

2. no memoryUsage set in activemq.xml
results:
- after couple thousand of messages casued Exception in activemq.log:
ERROR Queue - Failed to page in more queue messages 
java.lang.OutOfMemoryError: GC overhead limit exceeded

3. <memoryUsage limit="600mb"/> 
- any of previous problem

Isn't memoryUsage setting used only for non-persistent messages?
(http://activemq.apache.org/message-cursors.html). How can I predict how
much memory should I set for persisting couple GB of messages without having
broker in error state?
Furthermore, setting prefetch limit to 1000 leads to decrasing of persistent
store usage - why is that?

Thanks for your help.
Regerds,
Juraj














-- 
View this message in context: http://www.nabble.com/Problem-with-setting-memoryUsage-when-using-persistent-messaging-with-slow-consumers-tp23657227p23657227.html
Sent from the ActiveMQ - User mailing list archive at Nabble.com.