On Tue, 28 Mar 2000, Mark Hahn wrote:
> so how is your performance problem manifest?
When a massive buffer flush on my P3-500MHz, 1GB RAM server hits,
it slows down every NFS client and even the server's system console
to a crawl for 15-60 seconds.
When this happens, the ICP Vortex's GDT Monitor program shows it
being fed a constant write (e.g., it's buffers are 30-60% full
consistently). Yet, "top" still reports 95%+ idle.
I cannot figure out what the culprit is. One concern is the fact
that RedHat still ships with bdflush-1.5 and Caldera ships with
bdflush-1.62. And someone mentioned that bdflush has now been
replaced by kflushd (in the kernel?). I have kflushd, kupdate and
kswap all showing up in my ps/top list. /sbin/update reports that
it is from bdflush 1.4, although I have installed the 1.5 RPM.
Otherwise, I am trying to figure out how I can "tune" the cache
system to better accomodate the system. Right now, the system just
caches forever and writes extremely infrequently as far as I can
tell.
Again, any help at this point would be greatly appreciated. But I
cannot attempt "trail'n error" since it is a production system.
NFS v2 and just the load itself could be other culprits. But I
need some serious proof and understanding here. I'm just not
finding out what it is. What tools should I be using to find out?
Thanx, and sorry, I'm totally lost here.
-- TheBS
-- Bryan "TheBS" Smith -- Engineer, IT Professional and Hacker E-mail: mailto:thebs@theseus.com,b.j.smith@ieee.org Disclaimer: http://www.SmithConcepts.com/legal.html ************************************************************* TheBS ... Serving E-mail filters to /dev/null since 1989- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Fri Mar 31 2000 - 21:00:23 EST