From Bugzilla Helper: User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.0.0) Gecko/20020605 Description of problem: on my (dual 1GBram) Dell server doing stress compilation + Tweedee's mtest, vmstat fives bogus numbers for idle cpu Version-Release number of selected component (if applicable): How reproducible: Always Steps to Reproduce: 1.make -j4 in linux-2.4.18-4 (smp) 2.st's mtest running (mtest -m 700 -r 10 -w 3) 3.run vmstat Actual Results: yonkio 12:17 (96) rossetti/linux-2.4.18-4>vmstat 1 procs memory swap io system cpu r b w swpd free buff cache si so bi bo in cs us sy id ... 14 4 1 530064 8496 820 5220 344 0 396 0 135 105 84 17 21262213 Expected Results: reasonable numbers :) Additional info:
Created attachment 60269 [details] a longer vmstat run showing the problem...
I've also seen this on a Null beta install, when testing Apache 2.0 on a single 500Mhz PIII system. Linux xxx 2.4.18-12 #1 Fri Aug 16 16:57:26 EDT 2002 i686 unknown unknown GNU/Linux This usually happens a few times whilst running "ab -c 50 -n 100000 http://myhost/" from a different box.
Created attachment 73125 [details] Another sample of vmstat output
Looks like the high idle time is the result of integer overflow, since the sum of the system time and user time is greater than 100 in all the cases where it goes haywire.
Created attachment 76646 [details] fix for idle overflow
I've been experiencing this often also - can the fixed procps package be made available in the RH errata? It's a pain...
The fix is in 2.0.11 which is in rawhide and will be in the next release. If you want it in 7.3 i think rebuilding the SRPM will work fine.