[
https://issues.apache.org/jira/browse/DISPATCH-332?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Vishal Sharda updated DISPATCH-332:
-----------------------------------
Description:
We are running two Dispatch Routers each configured for interior mode and the
second router's configuration includes a connector to the first router.
When we connect one sender to one router and one receiver to the other router
both listening to the same queue, we see all messages (20,000 in our test)
being transmitted.
As soon as we start a second sender connected to the same router to which the
first sender connects and sending to the same queue, we start seeing heavy
message loss. Around 20% of messages are lost with each sender attempting to
send 20,000 messages on its own (40,000 in total) and running in parallel with
the other sender. The message loss happens regardless of the message size.
We tried with simple_send.py, simple_recv.py as well as send and recv C
executable files from Qpid Proton 0.12.2.
We even saw a crash in the router with the following message:
qdrouterd: /home/vsharda/qpid-dispatch/src/posix/threading.c:71:
sys_mutex_lock: Assertion `result == 0' failed.
Aborted
The message loss was observed with the 0.6.0 SNAPSHOT taken on May 9 as well as
the one taken on March 3 before the router core refactoring.
was:
We are running two Dispatch Routers each configured for inter-router mode and
the second router's configuration includes a connector to the first router.
When we connect one sender to one router and one receiver to the other router
both listening to the same queue, we see all messages (20,000 in our test)
being transmitted.
As soon as we start a second sender connected to the same router to which the
first sender connects and sending to the same queue, we start seeing heavy
message loss. Around 20% of messages are lost with each sender attempting to
send 20,000 messages on its own (40,000 in total) and running in parallel with
the other sender. The message loss happens regardless of the message size.
We tried with simple_send.py, simple_recv.py as well as send and recv C
executable files from Qpid Proton 0.12.2.
We even saw a crash in the router with the following message:
qdrouterd: /home/vsharda/qpid-dispatch/src/posix/threading.c:71:
sys_mutex_lock: Assertion `result == 0' failed.
Aborted
The message loss was observed with the 0.6.0 SNAPSHOT taken on May 9 as well as
the one taken on March 3 before the router core refactoring.
> Heavy message loss happening with 2 interconnected routers
> ----------------------------------------------------------
>
> Key: DISPATCH-332
> URL: https://issues.apache.org/jira/browse/DISPATCH-332
> Project: Qpid Dispatch
> Issue Type: Bug
> Components: Routing Engine
> Affects Versions: 0.6.0
> Environment: Debian 8.3, Qpid Proton 0.12.2 for drivers and
> dependency for Qpid Dispatch, Hardware: 2 CUPs, 15 GB RAM, 30 GB HDD.
> Reporter: Vishal Sharda
> Priority: Blocker
>
> We are running two Dispatch Routers each configured for interior mode and the
> second router's configuration includes a connector to the first router.
> When we connect one sender to one router and one receiver to the other router
> both listening to the same queue, we see all messages (20,000 in our test)
> being transmitted.
> As soon as we start a second sender connected to the same router to which the
> first sender connects and sending to the same queue, we start seeing heavy
> message loss. Around 20% of messages are lost with each sender attempting to
> send 20,000 messages on its own (40,000 in total) and running in parallel
> with the other sender. The message loss happens regardless of the message
> size.
> We tried with simple_send.py, simple_recv.py as well as send and recv C
> executable files from Qpid Proton 0.12.2.
> We even saw a crash in the router with the following message:
> qdrouterd: /home/vsharda/qpid-dispatch/src/posix/threading.c:71:
> sys_mutex_lock: Assertion `result == 0' failed.
> Aborted
> The message loss was observed with the 0.6.0 SNAPSHOT taken on May 9 as well
> as the one taken on March 3 before the router core refactoring.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
