Where can this be found? I am interested in trying this on my news
server.
> > ... I know its nasty what I'm
> > repeating ... but in the last few months there where a lot reports of
> > memory leaks ... IMHO the most of them haven't exist. One of the
> > real leaks was caused due an expire of an highly used news server ...
> > this leak is closed. Today the busmaster dma for trition causes a lot
> > of problems .. therefore it's save to let this optimisation off.
I am running 2.0.32 on my news server, and I still see what _appears_ to
be a leak. When the system has been up for a day (memory usage should
be fairly stabilized), there is 38M used not counting buffers/cache.
When it has been up for six days, after shutting down INN and all the
reader processes (leaving sendmail, cron, syslog/klog, gated, xntpd,
inetd, httpd, init, and 6 gettys running), there was 130M used! The
buffers/cache were not as large as they are after the system is up for
one day. The system starts slowing down after a few days, as it has
less and less memory availble for cache. The 'expire' process takes
longer every day, as there is less memory available for it to use.
I never saw any of this under 2.0.29 on my old news server. I would go
back to 2.0.29, except then I need a lot of patches for SYN floods,
teardrop, etc. I need to load a kernel with all of these fixes on my
other systems, but when I tried 2.0.31, I got the same kind of memory
leaks (after a few days I had to reboot or risk running out of memory
and crashing). Right now, we are evaluating whether to use Linux or
Solaris x86 for our next new server, and I want to run Linux. However,
my boss wants stability: our Solaris x86 box stays up for months, and
the brand new news server has to be rebooted every 5-6 days to keep it
running smooth.
I want to help figure this out! I don't have the knowledge or time to
do it by myself though. If ANYONE can help me, please send me email.
The reason I keep complaining on this list is because I still see
problems with every 2.0.x kernel above 2.0.29.
> It annoys me extreamly much that there is no practical way of detemine the
> amount of used kernel-memory. This is... Gnnnnh... F...... wrong!
Yeah, this is a major problem. That is why I thought I had a leak
before, because the memory used for buffer heads was not reported as
used anywhere. Maybe what I am seeing above is another case of
unreported use of memory. I don't think so, since it uses progressively
more and more RAM.
-- Chris Adams - cadams@ro.com System Administrator - Renaissance Internet Services I don't speak for anybody but myself - that's enough trouble.