Re: Parfait agent, docs, rpm updates

Hi,

On 2016-11-25 01:55, Nathan Scott wrote:
> 
>> Defining the same cluster ID for two different agents using different
>> name seems to work, it should probably fail to prevent accidents as
>> their PMID is the same in this case.
> 
> pmdammv is actually capable of stitching together metrics from multiple
> files - people even use this to provide different instances of a metric
> from different applications for example.  So, its not necessarily an
> accident when it happens, which makes guarding against accidents tricky.

Ok, perhaps a quick note on the man page might be helpful?

>> With WildFly (10.1.0.Final) I still see the same logging related
>> exception when trying to start it up after adding this to start script:
>>
>> JAVA_OPTS="-Djava.util.logging.manager=org.jboss.logmanager.LogManager
>> -Dparfait.name=wildfly -javaagent:/usr/share/java/parfait/parfait.jar
>> $JAVA_OPTS"
> 
> Not sure this is going to be fixable ... still looking for solutions, but
> its Wildfly being really paranoid about logging.  This may end up being a
> case where the proxy mode has to be used I guess, good thing we have that
> solution up our sleeve now.

After discussing with some Java folks here gave me impression that it's
not exceptional at all that JMX is used in production for metrics (after
all, the metric values don't appear out of thin air and in some cases
require the application to do some additional work making them available
so the cost of transportation (memory vs socket) to a client is not
necessarily the key factor), however in some cases JMX is not viable
(see below).

I will be offline for a few weeks so below is a braindump^W^W^W are some
ideas / topics that have come up around PCP/Java since our previous
correspondence, perhaps at least one them came up already:

- in order for PCP/Java being a compelling offer on this front a
complete set of metrics at least from the OpenJDK JVM should be
available and preferably from some of the most ubiquitous components
(WildFly, Cassandra, Tomcat, etc) as well

- there should be a couple of examples on how to add new metrics for
in-house applications

- BCI (byte code instrumentation) integration / support would be great
although I'm not yet completely sure what this actually means :-) A
trivial example could be that in R&D/QA BCI may be helpful to verify
class usage is as expected (so instead of the total class load count JVM
provides break that down to class level metrics). Byteman seems to enjoy
popularity on this front, not sure what it would take to for example
connect with it using Parfait or JSON or something lese. Or are there
some more elegant solutions that could be considered here for such cases.

Thanks,

-- 
Marko Myllynen

-=-=-=-=-=-=-=-=-=-=-=-
pcp mailing list
[email protected]
https://groups.io/g/pcp/messages
-=-=-
Groups.io Links:

You receive all messages sent to this group.

View/Reply Online (#14821): https://groups.io/g/pcp/message/14821
View All Messages In Topic (5): https://groups.io/g/pcp/topic/2984744
Mute This Topic: https://groups.io/mt/2984744?uid=174580
New Topic: https://groups.io/g/pcp/post

Change Your Subscription: https://groups.io/g/pcp/editsub?uid=174580

Group Home: https://groups.io/g/pcp
Contact Group Owner: [email protected]

Terms of Service: https://groups.io/static/tos

Unsubscribe: https://groups.io/g/pcp/leave/354243/563757577/xyzzy
-=-=-=-=-=-=-=-=-=-=-=-