SUMMARY: IOWAIT on a slice with no filesystem?

From: LA/ETH <>
Date: Fri Jan 14 2005 - 07:17:09 EST
OK, the answer was fairly quick and easy: I was looking at the wrong
place. Since the root disks are encapsulated, there is no point in
looking at the Solaris volumes, Veritas does not access the disks via
the c0t0d0sx interface. It is also useless to use iostat is such cases.
Answers are quoted below. Thanks to:

Sebastian Daubigne

for their quick and accurate answers

You have the OS mirrored via VxVM. The root disk c0t0d0 is encapsulated.
Let's look at WHAT VXVM does exactly:

Part      Tag    Flag     Cylinders         Size            Blocks
  0       root    wm   20624 - 21712        1.50GB    (1089/0/0)   3146121
  1       swap    wu   21713 - 24615        4.00GB    (2903/0/0)   8386767
  2     backup    wu       0 - 24619       33.92GB    (24620/0/0) 71127180
  3          -    wu       0 -     0        1.41MB    (1/0/0)         2889
  4          -    wu       1 - 24619       33.91GB    (24619/0/0) 71124291
  5 unassigned    wm    5083 - 20623       21.41GB    (15541/0/0) 44897949
  6        var    wm       1 -  1452        2.00GB    (1452/0/0)   4194828
  7        usr    wm    1453 -  5082        5.00GB    (3630/0/0)  10487070

In this layout, VxVM created Slice 3 as the private region and Slice 4 as
the data region.

When the OS tallks to any of these slice, VxVM talks to slice 4 at offset
??? to get all data. That is Not slice 0,1,5,6 or 7. As such you see the
overlap in slices. VxVM has it's own view of the disk once encapsulated and
does not use the normal S0,S1 ...

As such your I/O.

Maybe use vxstat to see what is REALLY going on, where the io is..


FS Partitions are not directly accessed when you have the rootdisk under
VxVM control.
Usually slice 4 is the the partition for the public region of VxVM.
The public region holds the data, so it is normal to see I/O activity on
this particular partition.

Now if you want to track the FS that drives the I/O activity, you can use
'vxstat' (e.g. "vxstat -v -i 5 -c 5").

And the original post: 

I have a Netra T1405 with 4 CPU, 4GB of RAM and 4GB swap running Solaris 8
(kernel Generic_108528-18). The machine runs Oracle OID (Oracle,
OID which stores its data on a T3. Oracle is put under Veritas
Cluster (3.4) control.
>From time to time, without any recognisable time or usage patterns the
machine experiences very high iowait which can only be stopped by restarting
the Oracle resource group. What makes it interesting for me is that iostat
(see printout below) seemes to indicate high disk busy rate on a slice of
the mirrored root disks which does not even have a file system on it, and of
course it's not the swap slice either.
sunmanagers mailing list
Received on Fri Jan 14 07:17:47 2005

This archive was generated by hypermail 2.1.8 : Thu Mar 03 2016 - 06:43:42 EST