I'd just like to second Peter's efforts, for several reasons, one of which
is I moved away from working with lustre awhile ago and currently don't
even have access to any of the types of configurations others like Peter
do, which is just one reason why there hasn't been any additional features
added to lustre support for many years. Furthermore, that fact that
Peter's company uses Lustre in their product line gives them a level if
expertise I may have once had but totally lost.
So what does this mean for current collectl/lustre users? In the short
term nothing will change. On startup collectl will look around and if it
see lustre is there, it will begin monitoring it using it's own internal
mechanisms. However, if you want to use Peter's plugin instead, you simply
import his module and collectl will use that instead.
Longer term, in say a year or so, assuming this new lustre module works to
everyone's satisfaction, I intend to completely remove all lustre code from
collectl itself, requiring everyone to use Peter's module so I highly
encourage everyone who currently uses collectl for Lustre monitoring give
it test drive and provide feedback to Peter.
My short term plan is release a new version of collectl but would like to
wait a couple of more weeks just to give people some time to shake out
Peter's code.
-mark
On Mon, Feb 17, 2014 at 4:33 PM, Peter Piela <[email protected]>wrote:
> Hello Everyone,
>
> my name is Peter Piela, and I work for a company named Terascala. We
> make high performance storage appliances based on the Lustre parallel
> filesystem. We have been using collectl to monitor Lustre for several
> years. During that time we have made several enhancements to the data
> collector, and added support for Lustre 2.x (tested up to 2.1.5). Last Fall
> I contacted Mark Seger about contributing our changes back to the
> community. He suggested that we make the Lustre data collector a separate
> plugin, and take ownership for its ongoing development.
>
>
>
> This plan made sense to us; initial development is complete, and we are
> looking for beta testers who are willing to try the plugins and provide
> feedback. Below are some high level notes describing the scope of the new
> plugins.
>
>
>
> · The Collectl Lustre plugins (lustreMDS, lustreOSS, and
> lustreClient) are based on the built-in Collectl 3.6.9 code
>
> · The plugins add support for Lustre 2.x (tested up to 2.1.5),
> and remove support for Lustre versions pre 1.8 and HPSFS
>
> · The plugins add new data collection options. Most of the new
> data elements are only available via Ganglia export; however, I am more
> than happy to work with users on extending the console reports and plots
>
> · The plugins are not compatible with the 3.6.9 version of
> gexpr.ph (Ganglia export). Our extensions to the Ganglia export mechanism
> will be incorporated into the next collectl release
>
>
>
> If you are interested in testing the Lustre plugins let me know, and I
> will provide you with software, and basic documentation that describes
> their usage.
>
>
>
> Peter
>
>
>
>
> ------------------------------------------------------------------------------
> Managing the Performance of Cloud-Based Applications
> Take advantage of what the Cloud has to offer - Avoid Common Pitfalls.
> Read the Whitepaper.
>
> http://pubads.g.doubleclick.net/gampad/clk?id=121054471&iu=/4140/ostg.clktrk
> _______________________________________________
> Collectl-interest mailing list
> [email protected]
> https://lists.sourceforge.net/lists/listinfo/collectl-interest
>
>
------------------------------------------------------------------------------
Managing the Performance of Cloud-Based Applications
Take advantage of what the Cloud has to offer - Avoid Common Pitfalls.
Read the Whitepaper.
http://pubads.g.doubleclick.net/gampad/clk?id=121054471&iu=/4140/ostg.clktrk
_______________________________________________
Collectl-interest mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/collectl-interest
