[
https://issues.apache.org/jira/browse/AMQ-7343?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Dominic Gagne updated AMQ-7343:
-------------------------------
Description:
We recently encountered an issue where a significant delay (often up to 15
seconds) occurs between the time a message is acknowledged by a consumer and
the time the next message is received from the broker. This issue only
manifests itself until fairly specific conditions, but we do have a
reproducible case. The problem occurs when the broker transport `nio+ssl` is
used and clients are running in Docker containers. ActiveMQ versions tested are
5.15.9 and 5.15.10. Default ActiveMQ broker configuration can be used (with the
exception of the nio+ssl transport). Disabling either nio or ssl in the broker
configuration prevents this issue from occurring.
Using Wireshark, we analyzed all ingress and egress traffic from the Docker
network while the consumer was running. This analysis revealed that there are
in fact long periods of no activity on the network, despite the fact that the
client immediately acknowledges the message and is ready to receive a new one.
We have been able to reproduce this on locally running brokers as well as
brokers hosted on AmazonMQ.
To reproduce using the attached example:
{{docker build -f ActiveMqTest/Dockerfile-ActiveMqTest -t 'activemqtest'
ActiveMqTest/}}
{{docker run activemqtest}}
was:
We recently encountered an issue where a significant delay (often up to 15
seconds) occurs between the time a message is acknowledged by a consumer and
the time the next message is received from the broker. This issue only
manifests itself until fairly specific conditions, but we do have a
reproducible case. The problem occurs when the broker transport `nio+ssl` is
used and clients are running in Docker containers. ActiveMQ versions tested are
5.15.9 and 5.15.10. Default ActiveMQ broker configuration can be used (with the
exception of the nio+ssl transport). Disabling either nio or ssl in the broker
configuration prevents this issue from occurring.
Using Wireshark, we analyzed all ingress and egress traffic from the Docker
network while the consumer was running. This analysis revealed that there are
in fact long periods of no activity on the network, despite the fact that the
client immediately acknowledges the message and is ready to receive a new one.
We have been able to reproduce this on locally running brokers as well as
brokers hosted on AmazonMQ.
To reproduce using the attached example:
{{docker build -f ActiveMqTest/Dockerfile-ActiveMqTest -t 'activemqtest'
ActiveMqTest/}}
{{docker run activemqtest}}
> Delayed Consumption of Messages
> -------------------------------
>
> Key: AMQ-7343
> URL: https://issues.apache.org/jira/browse/AMQ-7343
> Project: ActiveMQ
> Issue Type: Bug
> Components: Broker
> Affects Versions: 5.15.9, 5.15.10
> Environment: ActiveMQ version: 5.15.10 or 5.15.9
> ActiveMQ configuration: default
> Client: C# running in Docker container
> Transport: nio+ssl
>
>
> Reporter: Dominic Gagne
> Priority: Critical
> Attachments: ActiveMqTest_Consumption.zip
>
>
> We recently encountered an issue where a significant delay (often up to 15
> seconds) occurs between the time a message is acknowledged by a consumer and
> the time the next message is received from the broker. This issue only
> manifests itself until fairly specific conditions, but we do have a
> reproducible case. The problem occurs when the broker transport `nio+ssl` is
> used and clients are running in Docker containers. ActiveMQ versions tested
> are 5.15.9 and 5.15.10. Default ActiveMQ broker configuration can be used
> (with the exception of the nio+ssl transport). Disabling either nio or ssl in
> the broker configuration prevents this issue from occurring.
> Using Wireshark, we analyzed all ingress and egress traffic from the Docker
> network while the consumer was running. This analysis revealed that there are
> in fact long periods of no activity on the network, despite the fact that the
> client immediately acknowledges the message and is ready to receive a new one.
> We have been able to reproduce this on locally running brokers as well as
> brokers hosted on AmazonMQ.
>
> To reproduce using the attached example:
> {{docker build -f ActiveMqTest/Dockerfile-ActiveMqTest -t 'activemqtest'
> ActiveMqTest/}}
> {{docker run activemqtest}}
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
