[SUMMARY] High %wio, Why ?

My original post is below.

Thanks to:
 Shaw, Matthew 
 Kevin Buterbaugh 
 Casper Dik 
 Johan Hartzenberg 
 Terry Gardner
 Darren Dunham
 Bruce Kirkland 
 Mike Salehi

- This could be due to a stuck process, swapping and paging,
tape-backup, 
  or many other things. Try iostat -xncz 5 
  Let it run for a few iterations and look at the output.  Which disk is
  waiting on IO or have high busy or transaction times, is a clue to
what is
  working.

- The kernel updates a ticker every clock tick when a process is waiting

  for IO. Check your disks with iostat. Also, you can use ps -Lp <PID>
to 
  see which threads of a process are waiting. Then use pstack <PID> to
see 
  what the thread is doing.

- A %wio reading doesn't necessarily mean anything is wrong, especially
on
  this box which isn't really doing anything.  If you asked a tape drive

  to rewind, the command that does that will block until the drive
  rewinds, so it counts as io wait.  It doesn't mean that there's a
  problem.

- Wait I/O should be ignored if this is anything other than a single CPU
box.
  The reason is that once a CPU issues an I/O request, it is then free
to 
  process other waiting threads. If this is a single CPU box, then you
should 
  use "ps -ef" and look for sleeping processes (there'll be a bunch). 
  Look for the one that should be running, not sleeping.

- Not positive if it holds true in Solaris 8, but often NFS IO wait
times 
  are captured along with the regular io on the system which either more

  accurately represents or distorts the IO wait reporting on the system.

  (Depending on who you talk to)
  If you are indeed running nfs, I would check the output of nfsstat to
  see what your statistics are. Generally if you have a problem in these
  stats it will translate to high IO wait times for your box (If the NFS
  mounts are included when IOWAIT is calculated). This holds true even
  more so when you have running processes accessing IO from the nfs
  mounts. I have even seen bad connection between NFS client --> NFS
  server cause high IOWAIT times even when nothing is accessing the
mount.

Thanks all for the information, it turned out to be a bad disk ! 

Levi 
 


-----Original Message-----
From: sunmanagers-bounces@sunmanagers.org
[mailto:sunmanagers-bounces@sunmanagers.org] On Behalf Of Levi Ashcol
Sent: Tuesday, June 03, 2003 11:27 AM
To: sunmanagers@sunmanagers.org
Subject: High %wio, Why ?


Hi Gurus,
Below is the sar output for one of our Solaris 8/PL-15 server
15:19:33   %usr   %sys    %wio   %idle
15:19:35    0       0      0       98
15:19:35    0       0      40       60
15:19:35    0       0      40       60
15:19:35    0       0      40       60

I wonder why there is a high wio percentage ? As far as I understand 
this means that there is a certain process(es) waiting for I/O for a
Long period of time ! Am I right ? If so How can I find the process 
causing this waiting for the I/O high percentage ?

Any feedback will be appreciated, IWS.

Thanks

Levi
_______________________________________________
sunmanagers mailing list
sunmanagers@sunmanagers.org
http://www.sunmanagers.org/mailman/listinfo/sunmanagers
_______________________________________________
sunmanagers mailing list
sunmanagers@sunmanagers.org
http://www.sunmanagers.org/mailman/listinfo/sunmanagers