Hello, I am regularily looking at memory use of processes on solaris these days. Much of my work recently has been using libumem to track exactly what is using memory and it has proved very useful. I have found a limitation of using libumem in this way is that large allocations get classed as "oversize" and do not get tracked in libumem's caches so they do not appear in the output of libumem commands (except ::findleaks for which there is some special handling of oversize).
Could someone confirm that I am correct about this that oversize allocations are not tracked in libumem? It makes sense really. I am thinking it might be possible to extend libumem to allow tracking of oversize allocations. Would that be a relatively straight-forward task? I have the source >;) eek! To further our ability to test and debug what is the VERY best memory profiler tool on solaris? :) I have the feeling I am missing something? Am I? Thanks all for getting this far and bearing with yet another libumem email! James. aside: Attached is mdebug.sh script extended so it takes corefile on command-line. And extended with a -u option that runs libumem commands. e.g. mdebug.sh -u app core or mdebug.sh app core I find the ability to specify files on command-line much better. Simple changes but makes script much more usable - depending on your context maybe. -------------- next part -------------- An embedded and charset-unspecified text was scrubbed... Name: mdebug.sh URL: <http://mail.opensolaris.org/pipermail/mdb-discuss/attachments/20060728/88e77e7d/attachment.ksh>
