[ 
https://issues.apache.org/jira/browse/AMQ-5733?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14504276#comment-14504276
 ] 

Marc Schöchlin commented on AMQ-5733:
-------------------------------------

@James Green: i reviewed your problem and i agree

My intention was that a termination by "SIGKILL" should be something abnormal 
because the broker should shutdown in regular way.
SIGKILL does not provide the possibility to run shutdown logic (shutdown hook)  
for activemq - you should really consider to shutdown activemq by jmx.

Nevertheless according to LSB 
(http://refspecs.linuxbase.org/LSB_3.1.0/LSB-Core-generic/LSB-Core-generic/iniscrptact.html)
 the exit code should be 0 if activemq is really not running after running 
"/etc/init.d/activemq stop". 
I changed this and appended it to merge request 
https://github.com/apache/activemq/pull/61



> kill -9 always results in failure signal
> ----------------------------------------
>
>                 Key: AMQ-5733
>                 URL: https://issues.apache.org/jira/browse/AMQ-5733
>             Project: ActiveMQ
>          Issue Type: Bug
>          Components: Distribution
>    Affects Versions: 5.11.1
>            Reporter: James Green
>
> Using https://github.com/rgevaert/activemq2deb I am trying to distribute 
> upgrades to our fleet. This results in failure each time.
> Looking into the matter the activemq script tries to stop the broker via JMX. 
> This is not configured, so fails.
> It falls back to issuing "kill -9 <PID>" then immediately exits with code 1. 
> This code 1 causes the packaging script to bail and thus we end up with a 
> mess.
> A possible solution is to return the "kill" command's own result. This 
> doesn't necessarily reflect whether "kill" succeeded in terminating the PID 
> however.
> A second possible solution is to detect the presence of 'start-stop-daemon' 
> and use this in preference. Narrows down the cases in which it does not work 
> at least.
> I am of the opinion that a script aiming of cross-platform compatibility is 
> probably never going to work completely so detecting and making use of 
> per-system tooling is probably best long term, meantime would returning 
> "kill's" exit code help?



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

Reply via email to