Eric Schrock <eric.schrock at sun.com> said:
>>>
>>>On Fri, Jun 24, 2005 at 10:55:33PM +0100, Peter C. Tribble wrote:
>>>> 
>>>> I wrote a java kstat browser, and I'm tempted to do the same for prtpicl
>>>> (and maybe some of the other prt* commands), to make it easier to see the
>>>> hierarchy and drill down into it. It's hard to wade through prtconf and
>>>> prtpicl output (especially with -v on).
>>>> 
>>>
>>>On a further note, if you're willing to work on a complete libkstat JNI
>>>interface, this could be extremely beneficial to Solaris.  We're just
>>>starting to really elevate Java to a first class participant in systemic
>>>observability within Solaris.  We have the DTrace JNI nearly finished,
>>>and we're developing a SMF JNI interface as well.  A kstat interface
>>>would fill a very useful gap in this portfolio.

This is the one project I'm interested in doing right now. It's something
I've been thinking of for a while, and I've spent a reasonable amount of
time playing around with it. I can't lay claim to any great programming
expertise, but it wasn't that hard to get a functional interface.

So yes, I'm very interested in this and willing to work at it.

There is an existing implementation of a JNI interface to libkstat in
resource pools. The first decision would be whether to work with that or
develop a completely independent interface.

>>>I can imagine a tool which presents kstats in a heirachial manner on
>>>regular intervals, and the user can click on one the kstats and transfer
>>>to a DTrace view where we can see the kstats on a per-process basis.
>>>This would obviously only work for those kstats covered by the vminfo
>>>and sysinfo providers, but would be _incredibly_ cool, since it answers
>>>the first two questions any administrator tendds to have:
>>>
>>>     "What is happening to my system?"

Which often extends to a network of a hundred or more interrelated systems,
and it's the interactions between them that are interesting (hence my
interest in network observability). A fairly common case I see is where a
process on a client node has gotten into a wierd state (often a loop)
and is making odd requests to an NFS server. Sometimes it's making a
big (negative) impact on the fileserver; more often the server isn't
impacted but the easiest way to spot it happening on the network is to
look for the pattern on the server.

>>>     "Who is doing that to my system?"

Absolutely. Although after a little while you get to have a list of the
usual suspects ;-)

>>>Assuming that the problem is due to a particular process, of course.
>>>But you get the picture.



-Peter Tribble
MRC Rosalind Franklin Centre for Genomics Research
http://www.petertribble.co.uk/ - http://ptribble.blogspot.com/


Reply via email to