Pattern objectname #30
Conversation
| List<MBeanAttribute> getAttributesForMBean(Node mBean, String sampleDMax) | ||
| throws Exception { | ||
| String mBeanName = selectParameterFromNode(mBean, "name", null); | ||
| String mBeanPublishName = selectParameterFromNode(mBean, "pname", ""); |
There was a problem hiding this comment.
If the mBean name contained a wildcard and multiple MBeans matched the wildcard, we can't have all of them using an identical pname. Somehow, the pname needs to be changed for each. As suggested in my comment #3 (comment) in June, this can probably be implemented using regex groups. This involves two things:
- one or more wildcard(s) in the mBeanName would be wrapped in parentheses
- the
pnameattribute would contain some kind of placeholder for inserting the groups
The MessageFormat class may be suitable for expanding placeholders in pname.
Then the XML would contain something like:
<mbean name='org.hornetq:module=Core,type=Queue,address="jms.queue.(*)",name="jms.queue.*"' pname="queue_{0}">
<attribute name="MessageCount" type="int32" slope="both"/>
<attribute name="ConsumerCount" type="int32" slope="both"/>
</mbean>
and then if there were three queues with names like jms.queue.testA, jms.queue.testB and jms.queue.testC, the derived mBeanPublishName values, after replacement of the {0} placeholder would be queue_testA, queue_testB and queue_testC
Each mbean name given in the XML configuration file is expanded using the query method of MBeanServer. The query method takes in an ObjectName and returns all registered MBeans that matches it. If there is an error, for example a typo in the configuration, the malformed MBean isn't added to the MBeanSampler.
Properly formed patterns are expanded into ObjectName and queried individually.
|
@dpocock i looked at your suggestion and tweaked it a bit. |
Merge the HEAD of ganglia:master into this fix for #3.
Malformed patterns are ignored and valid patterns are expanded into
ObjectNames and added to the sampler one-by-one.