[
https://issues.apache.org/jira/browse/ARTEMIS-1934?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Robbie Gemmell updated ARTEMIS-1934:
------------------------------------
Description:
The broker can write out some connection data twice in certain circumstances,
corrupting the connection data, due to the way it handles the proton-j
transport output and sending it out to Netty. It fails to 'pop' the correct
amount of data from the proton transport that was actually transmitted,
under-accounting the amount, which results in some of the data being sent a
second time the next time output is generated and sent.
The underlying bug has been present for quite some time, but other changes from
2.6.0 onward change broker behaviour in ways that make it much more likely to
actually occur during real world usage.
was:The broker can write out some connection data twice in certain
circumstances, corrupting the connection data, due to the way it handles the
proton-j transport output and sending it out to Netty. It fails to 'pop' the
correct amount of data from the proton transport that was actually transmitted,
underaccounting the amount, which results in some of the data being sent a
second time the next time output is generated and sent.
> the broker sends duplicate AMQP connection data in certain conditions
> ---------------------------------------------------------------------
>
> Key: ARTEMIS-1934
> URL: https://issues.apache.org/jira/browse/ARTEMIS-1934
> Project: ActiveMQ Artemis
> Issue Type: Bug
> Components: AMQP
> Affects Versions: 2.5.0, 2.6.0, 2.6.1
> Reporter: Robbie Gemmell
> Assignee: Robbie Gemmell
> Priority: Critical
> Fix For: 2.6.2
>
>
> The broker can write out some connection data twice in certain circumstances,
> corrupting the connection data, due to the way it handles the proton-j
> transport output and sending it out to Netty. It fails to 'pop' the correct
> amount of data from the proton transport that was actually transmitted,
> under-accounting the amount, which results in some of the data being sent a
> second time the next time output is generated and sent.
> The underlying bug has been present for quite some time, but other changes
> from 2.6.0 onward change broker behaviour in ways that make it much more
> likely to actually occur during real world usage.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
