Re: RH 8; intermittend very high load
From: Stephan Wonczak (wonczak_at_Uni-Koeln.DE)
Date: 04/28/04
- Next message: users_at_linuxforums.org: "re:Ambit USB Cable Modem setup on Fedora Core 2 test 2"
- Previous message: users_at_linuxforums.org: "re:Static Routing"
- In reply to: Tommy Reynolds: "Re: RH 8; intermittend very high load"
- Next in thread: Tommy Reynolds: "Re: RH 8; intermittend very high load"
- Reply: Tommy Reynolds: "Re: RH 8; intermittend very high load"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Date: Wed, 28 Apr 2004 11:56:35 +0200
Tommy Reynolds wrote:
>
> On Tue, 27 Apr 2004 21:50:37 +0200, Stephan Wonczak wrote:
>
> > To me this sounds like something is wedged in 'D'-state; maybe an I/O
> > issue. The 2.4 VM is well-known for ... 'creative' behaviour.
> >
> > Look at the output of "ps -ealf | grep ' D '" and see if there are any
> > stuck processes (during the time of high load, of course :-) ). Also
> > looking at /proc/meminfo and /proc/slabinfo might provide clues.
>
> Those two "/proc" entries are good places to look for memory
> thrashing and/or swap issues.
>
> I don't get the connection with a process simply waiting in
> un-interruptable sleep causing a process scheduling backlog. As
> long as a process is on the un-interruptable wait queue the kernel
> doesn't even see the process anymore during schedule time; it does
> not get woken up for signal delivery or semaphore releases.
The OP said he had a load of 4 while the CPUs being more or less idle.
Usually this is an indication that a process is stuck in kernel space
somewhere. This does count towards the load average while actual CPU
consumption is zero. Wether this is the right thing to do was the
subject of a much discussion :-)
Anyway, in my experience this is usually caused by a process waiting
on IO; stuck NFS is a prime candidate for this (just try an 'ls' on a
NFS mount while the server is not responding. Result: Load 1, CPU 0.0%)
> Can you enlightenment about this?
I hope my point is a bit clearer now.
> No quibble about earlier 2.4 VM being buggy,
:-)
> but after about 2.4.14 this smoothed out pretty well.
Yes, the overhaul of the 2.4 VM in the 2.4.10 timeframe was needed,
but is still is not too good for some loads. Especially on big memory
machines (...as I learned the hard way.... *sigh*) Some of the problems
have been sorted out, but we still have not reached a really stable
environment on outr big machines.
> HEY! ORIGINAL POSTER! What does the 'uname -r' program say?
Yes, and the outpuit of /proc/meminfo and /proc/slabinfo would be
helpful, too!
-- C U, Stephan "I haven't lost my mind; I know exactly where I left it." "The meaning of my life is to make me crazy"
- Next message: users_at_linuxforums.org: "re:Ambit USB Cable Modem setup on Fedora Core 2 test 2"
- Previous message: users_at_linuxforums.org: "re:Static Routing"
- In reply to: Tommy Reynolds: "Re: RH 8; intermittend very high load"
- Next in thread: Tommy Reynolds: "Re: RH 8; intermittend very high load"
- Reply: Tommy Reynolds: "Re: RH 8; intermittend very high load"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Relevant Pages
|