Thanks for any info,
Mark (relative HPUX admin novice)
> Can anyone explain to me what vxfsd is?
vxfs daemon...it handles lowlevel filesystem functions. It is busy
when there is a lot of disk activty.
> "top" shows it as always in run
> state. I'm trying to track down the cause of a poorly performing system
> and this is the prime suspect for now, but I've been unable to dig up any
> information about it yet. My load averages are always greater than 1,
> even when it's the wee hours of the morning and nobody else is logged in
> and all I'm doing is running top.
Unfortunately, classic Unix tools make it difficult to locate the
real problem. top shows compute-bound processes but won't help
with LAN I/O or disk I/O. Load a copy of Glance from your Application
CDs and you'll find the problem(s) a lot faster.
--
?
/'^'\
( o o )
*====oOOO===(_)===OOOo======*=====================================*
| Bill Hassell | Hewlett-Packard Response Center |
| SysAdmin Oooo. | b...@atl.hp.com / Atlanta, GA. |
*======.oooO===( )========*=====================================*
( ) ) / "There are two types of computer users in the
\ ( (_/ world...those that have lost data, and those
\_) that are going to." (blh, circa 1972)
I second that motion ...
Glance or Glance Plus is one of the only ways to really find out what is
happening ...
Ive.
i see it in run state all the time as well, even with no disk activity,
so load average is >=1. i had guessed it ran at some high priority wait
that was counted in load average. BSD Unixes will do this.
Hm. Maybe it's just a slow machine, and all is right with the world.
:-)
A good first question to ask is how exactly are you measuring a poorly
performing system?
Why was vxfsd the prime suspect? ie simply because it was using CPU, or are you
suspecting an IO bottleneck of some type, and if so then why?
Eric Stahl
>Mark Slagell m...@iastate.edu
>Can anyone explain to me what vxfsd is?
>"top" shows it as always in run
>state. I'm trying to track down the cause
>of a poorly performing system
>and this is the prime suspect for now,
<SNIP>
Hi Mark,
I'm afraid that any mutlithread process will always show up in the run
state and HP-UX 11. Both ps and top show this. It's just because they
don't understand what the process is doing. Only Glance (as others have
pointed out) make any sense of these jobs.
This should not be affecting your load average as none of vxfsd's
threads should be running. The load averages are the effective length
of the run queues, and these work at the thread not process level.
Do a "ps -el" and look at what else has a priority in the range
128->153.
Cheers
Ken