[
https://issues.apache.org/jira/browse/ARTEMIS-3148?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17292181#comment-17292181
]
Terrien Jean-Yves commented on ARTEMIS-3148:
--------------------------------------------
My problem is not to migrate data from mkahadb.
With only one log for all queues, if a message is not consumed on a queue, the
space occupied on disk by all messages from other queues is not freed even if
they are consumed.
With activeMQ to solve this problem the solution is to use mkahadb which
defines 1 base per file. thus blocking a message on one queue does not impact
the others.
With Artemis it seems like you can't have one journal per queue. therefore if 1
message is not consumed, the journal cannot be freed, which means that all the
messages from the other queues are not freed. with 10,000 queues it is very
common to have a queue that is suspended. and as they brew millions of
messages, no more journal files can be released which explodes the disk space
used.
How does Artemis solve this problem?
> ActiveMQ Artemis mkahadb migration
> ----------------------------------
>
> Key: ARTEMIS-3148
> URL: https://issues.apache.org/jira/browse/ARTEMIS-3148
> Project: ActiveMQ Artemis
> Issue Type: Improvement
> Components: ActiveMQ-Artemis-Native
> Environment: IA64 HP-UX (actual and possible target for artemis)
> X64 ReadHat Linux (Possible target for artemis)
> 11 ActiveMQ with 10 000 persisted queues ~1000000 messages per day
> I can't post on the list anymore
> http://activemq.2283324.n4.nabble.com/ActiveMQ-User-f2341805.html
> Reporter: Terrien Jean-Yves
> Assignee: Clebert Suconic
> Priority: Major
>
> Hello
> We have a system with ActiveMQ which manages a few thousand queues.
> We were faced with a dificulty. If on an instance of ActiveMQ 1 persisted
> queue does not have a consumer, one message is enough in this queue for the
> size of the KahaDB database to increase significantly. Indeed a message not
> consumed in one of the queues of the base prevents the system from freeing
> space for all queues on this database. And we are hiding millions of
> messages. Except with so many tails this is a common situation. To avoid this
> we used mkahaDB. Thus only the queue without a consumer is impacted. The
> other queues can free up space.
> We are considering switching to Artemis. But I don't find any information on
> this specific point in the documentation.
> Won't the new journal clog the filesystem?
> Where can I find the documentation to ensure that in the event of a consumer
> suspension, all messages consumed in other queues will be deleted.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
