[Red5] Graphing Red5 again

[Walter Tak]

>>> So i've investigated in 3 ways to possible graph the jvm data in
>>> red5.
>>>
>>> One way was to just expose the standard SNMP information, it cannot
>>> access any of the red5 mbeans however ie graphing connected
>>> clients. I
>>> have 3 graphs kinda working from an example cacti jvm graph template
>>> but still not working properly really as I would have expected.
>> ...
>>
>> I'm using rrdtool to make graphs on my network stats. Can you get
>> the stats
>> you need out of the running java processes using default commandline-
>> tools
>> or is that the core of the problem ?
>>
>
> We already have SNMP collecting everything available to the operating
> system and graphing in cacti, i want to expose memory , heap, thread,
> GC stuff, aswell as some red5 mbeans. What is your network stats in ?

Not sure what you mean with this: "What is your network stats in ?"

You mean integers ? bytes ?

I'm collecting input-output numbers from iptables, including some tables 
that filter incoming and outgoing port 1935 traffic so I know
how much real rtmp traffic the Red5 server is doing.

All specific Java numbers are interesting for debugging idd. If you can 
extract them without difficult SNMP integration than
just write a simple bash-script that adds the data to rrdtool.

SNMP is great if you want to monitor your server using a single centralised 
tool that can create trees of servers etc. But I doubt those
managementtools (I used one of HP many years ago, Openview)  can handle and 
display those very specific java-figures.

Most Red5 servers will be hosted as a single server or in a small cluster at 
'some' ISP/datacentre, outside LANs and WANs where
unified servers are running and have to be maintained by sysadmins.

Thus a simple solution would be most wanted by most of us Red5-server 
developers (and admins in one, often).

> Looks like im getting somewhere with this, but opennms is poorly
> documented., the graphs are not showing anything yet.
>
> I had a way of exposing some SNMP data from the red5 jvm, but i wasnt
> sure which data to use as the datasource for rrd. Maybe you might shed
> some light what useful bits of data is good ?

Well gl. I hacked mibs and oids like [.1.3.6.1.4.1.9237.2.1.1.1.3] ages ago 
and I rather stay away from that :)

I'd rather create a few more lines in a bash-script and feed the rrdtool 
database and then make a nice graph using "rrdtool graph".

That's way more transparent, easier to modify and only requires a script to 
extract the nescessary data and a default install of rrdtool to collect and 
graph it.

Anyone ontopic ; are you trying something like this perl script : 
http://prefetch.net/code/jvmstats.pl ?

#   jvmstats-gather.pl retrieves performance metrics from the
#   Java SNMP agent, and writes these entries to a file. This
#   file can then be analyzed by ORCA to produce several
#   performance graphs.
Regards,
Walter