incubator-kato-spec mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Steve Poole <>
Subject Re: Snapshot Application development usecases
Date Tue, 05 Jan 2010 10:33:17 GMT
On Tue, Jan 5, 2010 at 9:38 AM, David Griffiths

> Hi Steve, thanks for the clarification. A lot of the issues we assist
> customers with turn out to be problems in third-party code and for
> these problems a post-mortem dump of some sort is the starting point.
> I'm surprised there's so little interest in obtaining Java info from
> core dumps. A lot of our customers use Java thread dumps and heapdumps
> to solve their own problems and I'm sure they would do more with core
> dumps if the tooling hadn't in the past been so unwieldy (eg
> jextract).
> In all the discussions we've had with developers the first hurdle has
always been  getting them to understand that there is value to* them*  in
using dumps.
The Java developer community has grown up without the dump concept  (If you
do get a developer to acknowledge the existence of dumps then they see them
as either an opaque thing to be given to some other support team or as a
data collection mechanism for profiling)

Once you do get the  dump idea across you then have to explain that
actually, the most prevalent dump (ie HPROF)  doesn't have all the data
needed to help solve the problems developers want to see solved.    Which is
where this project comes in -  we need new dump artifacts that can help fill
in the missing gaps and which work with the Sun JVM.   We had hoped to get a
corefile reader that could inspect the Sun JVM stuctures etc  but that has
not happened ( and as I explained it not something we can do here)

Anyway, my perspective on all this is sufficiently different that I'll
> try to refrain from blundering in with any more dumb comments :)
I don't think you made any  dumb  comments - but perhaps your perspective
was initially too IBM centric.    We do welcome any comments ,  suggestions
, contributions etc  on any part of this project.
I don't want to scare you off so if something is unclear just ask away and
thanks for taking time to comment in the first place.

> Cheers,
> Dave


  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message