[
https://issues.apache.org/activemq/browse/AMQNET-241?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Ted Carroll updated AMQNET-241:
-------------------------------
Attachment: NmsRepro_clean.zip
Added a program that reproduces the issue.
To answer the previous question, what I'd like to be able to do is return the
error to the user and, when the next request comes in, be able to still use the
connection. If ActiveMQ comes back up in the meantime, I would like the
connection to work.
Right now we have two choices. 1) wrap all calls with try catch blocks and
tear down associated producers, consumers, sessions, session pools, and the
connection or 2) Remove the max retry limit and appear to be "hung".
Thanks,
Ted C.
> NMS Failover causes CPU utilization to spike if ActiveMQ is stopped.
> --------------------------------------------------------------------
>
> Key: AMQNET-241
> URL: https://issues.apache.org/activemq/browse/AMQNET-241
> Project: ActiveMQ .Net
> Issue Type: Bug
> Components: ActiveMQ
> Affects Versions: 1.2.0
> Environment: .NET 3.5
> Reporter: Ted Carroll
> Assignee: Timothy Bish
> Fix For: 1.2.1, 1.3.0
>
> Attachments: NMS_Failover_CPU_Spike1.patch, NmsRepro_clean.zip
>
>
> 1) Configure a simple program that attempts to connect to ActiveMQ with the
> url "activemq:failover:(tcp://localhost:61616)"
> 2) Make sure activemq is not running on the local host
> 3) Try to connect.
> 4) Notice that the CPU utilization in the sample program is high.
> 5) Start ActiveMQ.
> 6) Notice that the connection is successful but the CPU utilization remains
> high.
> Expected:
> I expect, potentially, an initial spike in CPU utilization but not sustained
> high CPU.
> Diagnosis:
> In FailoverTransport.DoConnect() there is the following if statement:
> if(ConnectedTransport != null || disposed ||
> connectionFailure != null)
> {
> return false;
> }
> else
> it apears that connectionFailure is set and never reset. Then the
> FailoverTask.Iterate() calls into DoConnect only to return false immediately
> every time.
> I have attached a patch, but I would be uncomfortable calling it a long term
> solution as I've not had the time to fully understand the code in question.
> However, it seems to improve the behavior.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
