On Wednesday 16 July 2008, Daniel Barowy wrote: > Hello everyone, > > Can someone here shed some light on why my vmstat might have > negative values? If I understand the manpage correctly (and Henning > and Phillipp's paper on tuning), these values are nonsensical: > > $ vmstat wd0 wd1 sd0 > procs memory page disks traps > cpu > r b w avm fre flt re pi po fr sr wd0 wd1 sd0 int > sys cs us sy id > 0 3 0 64360 16776 23 0 0 0 0 0 -14 -0 -0 43 > 4294967211 22 -1 -1 102 > > I'm comparing these values (from a very unbusy system) to a very > busy system, but I didn't think it would be negative-busy ;^) > > This is on an i386 OpenBSD 4.0 system. > > Thanks, > Dan
First of all, you need to upgrade to 4.3 rather than letting your system languish until no one remembers its idiosyncrasies anymore, and no one who stays up to date can attempt to replicate your problem. Secondly, in the 4.2->4.3 changes, you'll see that vmstat(8) had some very significant work done to it. http://www.openbsd.org/plus43.html In short, *my* vmstat on 4.3-STABLE i386 works a lot differently than *your* vmstat on 4.0 i386. Yep, I even mounted a wd0 wd1 and sd0 on my system to see if I could replicate your results. If you can replicate the problem on a supported system, preferably 4.3 or better, a -CURRENT snapshot, then it's actually a problem, but otherwise, it's a self inflicted wound caused by a failure to upgrade. Kind Regards, JCR
