First of all thanks to the following people for there help full a copy of
everything they sent to me is available just me e-mail for it
Jan.Berger.Henriksen
Hellebo Knut Knut.Hellebo_at_nho.hydro.com
brock
Nick Hill - RAL CISD VMS Systems
Martyn Johnson maj_at_cl.cam.ac.uk
Steve Mclaughl
John P. Speno, speno_at_swarthmore.edu
Scott Stevens s.k.stevens_at_ic.ac.uk
Michel Cyr Michel.Cyr_at_sidoci.qc.ca
Dave Cherkus
alan alan_at_nabeth.cxo.dec.com
Alex M. George
Allan Small (Cafeine propelled)
John Richards
Next a copy of the original problem
Ever since I've upgraded to Digital UNIX v3.2 I've been experiencing
problems with machines crawling along.
At first I thought this was related to linking, but I've just re-installed
V3.2 on an Alpha 2100, I had previously upgraded but had a few minor
problems and went for a full install. Ever since the machine will go
incredibly slow for a few minutes before returning to normal.
I've loaded up monitor and this tells me that 99% of the CPU time is being
allocated to sys. This would explain it going slow, so I've loaded up top
but apart from telling me the average load is around 20 (!) there's no
report of any process taking more than 5% of the CPU time.
I'm now at a loss to know how to track it down further, so my questions are
1) Any ideas on how to track what's stealing all the cpu time.
2) Anyone else got a similar problem, or is it a known problem.
Perhaps I should mention none of the users are doing anything extra to the
work form before, so I'm assuming in must be Operating System related.
----------------------------------------------------------------------------
------------------------------------------------
Now various people pointed out that I should use ps aux to get the top
process (I really will have to read the man pages on ps thoroughly there are
so many options I don't know about.). This confirmed that the kernel was
taking a massive whack of the CPU time. Dave Cherkes suggested vmstst -M
would show if any kernel process were taking a large amount of Memory, none
seemed to be too far out of line. Neither was the machine swapping, nor the
disk queues to high nor the number of system calls excessively large (
Monitor is a really useful program as it shows all of this on one screen -
very handy when your machine takes a good couple on mins to return the
results of a command)..
In the end I tried cutting down the amount of shared memory I'd defined (I'm
using a very high setting) and that seems to have done the trick.
One thing several people commented on was that 3.2 does seem to have
problems running in 32Mb of Memory (not a problem for me)
I do still get the same symptoms (all the cpu being stolen by sys) when
linking but I get this on all my 3.2 machines. Digital tech. support tell
me I'm not the only one to get this problem with linking but it's connected
with shared libraries. If anyone else has got any more information on this
then I'd like to hear from them (I'll summarize)
Simon Monday
mondays_at_logica.com
TEL +44 1372 227046
FAX +44 1372 227009
Received on Wed Jul 26 1995 - 12:22:00 NZST