[
https://issues.apache.org/jira/browse/ARTEMIS-4781?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17850225#comment-17850225
]
Erwin Dondorp commented on ARTEMIS-4781:
----------------------------------------
> Is the "simple extra consumer on the address ExpiryQueue" receiving and
> acknowledging those (now large) messages?
that one is a simple JMS consumer which uses autoconfirm to acknowledge any
message that it reads.
the messages in data/large-messages directory remained even after all clients
were stopped.
> Please clarify
Can I start over again?
After more testing, I found that files were left in data/large-messages when
sending AMQP messages with a bytes-payload between 99304 and 99704 bytes. I
tested in increments of 100 bytes, starting from 96KiB. These are exactly the
messages that, on creation, stay below my setting of
amqpMinLargeMessageSize=100000; and which after expiry+transfer became slightly
larger than that. It still only happens when the messages are also transferred
between nodes on expiry. i.e. it does not happen on single node installation.
> on-disk files for large messages are not always removed on expiry
> -----------------------------------------------------------------
>
> Key: ARTEMIS-4781
> URL: https://issues.apache.org/jira/browse/ARTEMIS-4781
> Project: ActiveMQ Artemis
> Issue Type: Bug
> Components: Clustering
> Affects Versions: 2.33.0
> Reporter: Erwin Dondorp
> Priority: Major
>
> SETUP:
> Using a broker-cluster.
> The tests are executed with durable and non-durable messages. 3 durable
> messages and 3 non-durable messages are produced every 60 seconds (almost) at
> the same time on the 1st broker.
> We are producing large AMQP messages and leave them on a durable queue.
> MSG/TMP files are created in directory `large-messages` for this as expected.
> After the configured amount of time, the messages expire as expected. the
> original MSG/TMP files are removed as expected.
> For monitoring, we have an simple extra consumer on the address `ExpiryQueue`
> connected to a 2nd broker in the same cluster.
> OBSERVATION:
> The MSG/TMP files are left on the disk of the 2nd broker also every 60
> seconds. This is unexpected. No related logfile lines are seen on either
> broker.
> The content of the MSG/TMP files is (based on it size) related to the
> original MSG/TMP files. These files have different names, likely because they
> have been recreated in the context of the ExpiryQueue address. The files are
> slightly larger, likely because of the addition of a few expiry related
> headers.
> Note that this also happens for messages that originally were just a few
> bytes too small to become a large message, but have become large messages due
> to the expiry process.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@activemq.apache.org
For additional commands, e-mail: issues-h...@activemq.apache.org
For further information, visit: https://activemq.apache.org/contact
