qpid-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From 李翰林 (JIRA) <j...@apache.org>
Subject [jira] [Updated] (QPID-6676) QPID 0.32 throughput drop sharply with HA module
Date Tue, 04 Aug 2015 07:12:04 GMT

     [ https://issues.apache.org/jira/browse/QPID-6676?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]

李翰林 updated QPID-6676:
----------------------
    Description: 
I build QPID 0.32 C++ broker with HA module on CentOS6.5 and run a cluster of 2 nodes with
--ha-replicate=all, Then send messages using QPID Messaging API continuously with protocol
AMQP0-10 and I meet a strange problem,If I purge all of messages in a queue and continue to
send messages,The throughput drop sharply. And the same problem seen on QPID 0.34

Steps to reproduce:
1:
    qpid-config add exchange direct testHA.exchange
    qpid-config add queue testHA.queue --max-queue-size 1048576000
    qpid-config bind testHA.exchange testHA.queue testHA.binding
2:
    Run "qpid-queue-stats -f testHA.queue" on a console
3:
    Send 100000 messages to testHA.exchange with routingkey testHA.binding
4:
    Purge queue use the QPID manager tool (qpid-tool)
5:
    Repeat step 3

In the step 3, I can send about 8000 messages per second one producer, and In step 4, I purge
all messages in the queue, then send messages continue, I can only send about 300 messages
per second, But when I shutdown the backup node the throughput back to normal



  was:
I build QPID 0.32 C++ broker with HA module on CentOS6.5 and run a cluster of 2 nodes with
--ha-replicate=all, Then send messages using QPID Messaging API continuously with protocol
AMQP0-10 and I meet a strange problem,If I purge all of messages in a queue and continue to
send messages,The throughput drop sharply. And the same problem seen on QPID 0.34

Steps to reproduce:
1:
    qpid-config add exchange direct testHA.exchange
    qpid-config add queue testHA.queue --max-queue-size 1048576000
    qpid-config bind testHA.exchange testHA.queue testHA.binding
2:
    Run "qpid-queue-stats -f testHA.queue" on a console
3:
    Send 100000 messages to testHA.exchange with routingkey testHA.binding
4:
    Purge queue use the QPID manager tool (qpid-tool)
5:
    Repeat step 3

get the output from step 2:
Queue Name          Sec        Depth    Enq Rate    Deq Rate
===============================================
testHA.queue  1064.00         11350         10.67            0.00       <== step3
testHA.queue        2.00         28965     8807.35            0.00
testHA.queue        2.00         46650     8842.57            0.00
testHA.queue        2.00         64300     8825.08            0.00
testHA.queue        2.00         81950     8824.85            0.00
testHA.queue        2.00         99600     8825.04            0.00
testHA.queue        2.00       100000       199.97            0.00
testHA.queue      40.01                 0           0.00      2499.65       <== step4
testHA.queue      16.00             250         15.63            0.00       <== step5
testHA.queue        2.00             900       325.07            0.00
testHA.queue        2.00           1600       350.05            0.00
testHA.queue        2.00           2300       350.12            0.00
testHA.queue        2.00           3000       349.98            0.00
testHA.queue       2.00            3700       349.96            0.00
testHA.queue       2.00            4350       324.96            0.00





> QPID 0.32 throughput drop sharply with HA module
> ------------------------------------------------
>
>                 Key: QPID-6676
>                 URL: https://issues.apache.org/jira/browse/QPID-6676
>             Project: Qpid
>          Issue Type: Bug
>          Components: C++ Broker, C++ Clustering
>    Affects Versions: 0.32, qpid-cpp-0.34
>         Environment: CentOS6.5 X86_64
>            Reporter: 李翰林
>            Assignee: Alan Conway
>            Priority: Critical
>
> I build QPID 0.32 C++ broker with HA module on CentOS6.5 and run a cluster of 2 nodes
with --ha-replicate=all, Then send messages using QPID Messaging API continuously with protocol
AMQP0-10 and I meet a strange problem,If I purge all of messages in a queue and continue to
send messages,The throughput drop sharply. And the same problem seen on QPID 0.34
> Steps to reproduce:
> 1:
>     qpid-config add exchange direct testHA.exchange
>     qpid-config add queue testHA.queue --max-queue-size 1048576000
>     qpid-config bind testHA.exchange testHA.queue testHA.binding
> 2:
>     Run "qpid-queue-stats -f testHA.queue" on a console
> 3:
>     Send 100000 messages to testHA.exchange with routingkey testHA.binding
> 4:
>     Purge queue use the QPID manager tool (qpid-tool)
> 5:
>     Repeat step 3
> In the step 3, I can send about 8000 messages per second one producer, and In step 4,
I purge all messages in the queue, then send messages continue, I can only send about 300
messages per second, But when I shutdown the backup node the throughput back to normal



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@qpid.apache.org
For additional commands, e-mail: dev-help@qpid.apache.org


Mime
View raw message