Collectl V3.3.1 just released

[MarkSeger]

I don't have a good handle on how many people who participate in this forum use collectl, but I wanted to let people know I just released a new version, which has 2 pretty cool features.
- it can send its output to ganglia
- you can now add your own data sources

This has been testing on a 2000+ node cluster, taking samples every 10 seconds so it clearly scales.

-mark

[Ivan Ferreira]

Thank you Mark. I can't imagine my life without collectl ;). It's a great tool. You are doing a great job and contribution.

I don't know why is not distributed with the media yet.

[MarkSeger]

glad you're finding it useful. As for distribution, it IS part of Fedora and perhaps some day will make it into RHEL. IT's also included in a few other distros as well.

As a side-topic, are you or others planning on attending the HP Tech Forum in June? I'll be doing a presentation on Advanced collectl topics. shameless plug... ;-)

-mark

[Steven McCoy]

Collectl is a great tool, thanks for all your hard work, Mark.

[Court Campbell]

> -you can now add your own data sources

I have been using collectl for little over a year, and this is definitely a cool new feature. Now if you can only convince the HPUX folks to implement procfs.

[MarkSeger]

court - can I assume by your comment you're going to try out the API? I'm definitely looking for feedback. I certainly have no problems with it but I also know collectl pretty well.

If you look at hello.ph and misc.ph, those took on the order of an hour each, give or take, and most of that was consumed by testing all the different output formats, getting columns lined up properly, making sure the write bits went over the socket and plot format was correctly populating the 'tab' file. And then there's making sure time formatting works correctly too. Maybe one or two other things as well.

The easiest way to get started is rename one of the existing scripts, come up with what I'm calling a 'key' and start hacking away. If you do stumble, just post some code and I'll walk you through it.

The one caveat is it's really easy to write inefficient code (I've done that a lot) and you need to be careful not to kill collectl's low-overhead.

-mark

[Court Campbell]

I'll give it a try. The only issue I am having is trying to figure out what to add. Collectl already does so much that is is hard to come up with some new data source.