[
https://issues.apache.org/jira/browse/KARAF-4850?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Jean-Baptiste Onofré resolved KARAF-4850.
-----------------------------------------
Resolution: Fixed
> Be able to specify several object names for the JMX collector
> -------------------------------------------------------------
>
> Key: KARAF-4850
> URL: https://issues.apache.org/jira/browse/KARAF-4850
> Project: Karaf
> Issue Type: Improvement
> Components: decanter
> Affects Versions: decanter-1.2.0, decanter-1.3.0
> Reporter: Vincent Zurczak
> Assignee: Jean-Baptiste Onofré
> Priority: Minor
> Fix For: decanter-1.4.0
>
>
> When using the JMX collector, there is a property called *object.name*, which
> can be a query for MBeans. However, the syntax of ObjectName is too poor IMO.
> And gathering all the MBeans information can result in sending too much data
> to the appender (e.g. ElasticSearch).
> It would be very convenient if one could customize which MBeans are retrieved
> exactly, as a list, in the same configuration file.
> Example:
> {quote}
> object.name = \
> java.lang:type=Memory,name=HeapMemoryUsage |\
> java.lang:type=Memory,name=NonHeapMemoryUsage |\
> java.lang:type=OperatingSystem,name=SystemLoadAverage |\
> java.lang:type=OperatingSystem,name=SystemCpuLoad |\
> java.lang:type=OperatingSystem,name=ProcessCpuLoad |\
> java.lang:type=OperatingSystem,name=FreePhysicalMemorySize |\
> java.lang:type=Threading,name=ThreadCount
> {quote}
> I suggest to use the pipe character as a separator.
> Or, we could use the base property as a prefix.
> {quote}
> object.name.1=java.lang:type=Memory,name=HeapMemoryUsage
> object.name.2=java.lang:type=Memory,name=NonHeapMemoryUsage
> object.name.3=java.lang:type=OperatingSystem,name=SystemLoadAverage
> object.name.4=java.lang:type=OperatingSystem,name=SystemCpuLoad
> object.name.5=java.lang:type=OperatingSystem,name=ProcessCpuLoad
> object.name.6=java.lang:type=OperatingSystem,name=FreePhysicalMemorySize
> object.name.7=java.lang:type=Threading,name=ThreadCount
> {quote}
> In terms of performance, it should not have a major impact with respect to
> the current implementation. We need to parse the list of values and iterate
> over it to fullfill the set of ObjectNames. Obviously, this could be achieved
> by creating several CFG files. But you would have to copy the JMX settings
> over and over again. IMO, it is better to have one configuration file per JMX
> configuration.
> I could work on it and submit a patch if necessary.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
