VXVM : Read error on large VxFS filesystem

[Q4you]

The following vxvm filesystem (lvol1, striped to 9 luns, on 9 different Fiber chan cards on Xp512 array) is setup for oracle dumps like this ->

dg rootdg 4 default 0 985293812.1025

dm disk01 c45t1d3 simple 1024 14225328 NOHOTUSE
dm disk02 c31t9d4 simple 1024 14225328 NOHOTUSE
dm disk03 c47t1d5 simple 1024 14225328 NOHOTUSE
dm disk04 c47t9d5 simple 1024 14225328 NOHOTUSE
dm disk05 c43t9d3 simple 1024 14225328 NOHOTUSE
dm disk10 c41t9d2 simple 1024 14225328 -
dm disk11 c37t9d1 simple 1024 14225328 -
dm disk12 c39t1d2 simple 1024 14225328 -
dm disk13 c35t1d1 simple 1024 14225328 -

Some of the disks in rootdg are not "nohotuse" and some are..Does it make any difference in the "vxvm" functioning ? ( The rootdg has no root file systems)

Oracle suggests that the problem may be a "logical read issue" between the
disks and the server.

The file system is mounted at /XYZ with ->

/dev/vx/dsk/rootdg/lvol1 /XYZ vxfs delaylog,nodatainlog,mincache=direct,convos
ync=direct 0 2

-TIA

Q

[Mark Travis]

Well "read error" is extremely vague. Can you be specific?

And "NOHOTUSE" shouldn't matter. That flag defines whether or not the disk is used in hot relocation.

First, middle, last:

Call Veritas' tech support. They will walk you through this. You do have phone support with Veritas, right? Don't mess around with this. Especially if this is a production machine.

If you don't have paid tech support with Veritas and can't call them (and your menagement doesn't approve T&M) then please describe your problem in more detail. Folks on this list ought to be able to get you close to a solution.

[Steven E. Protter]

Sounds like someone pulled a disk on the disk array.

That absolutely should not cause any issues like you are receiving.

I would fear a major problem. I would call HP support for a critical issue and be ready to shut that database down and back it up.

For OLTP Oracle does not reccomend your setup, they actually recommend Raid 1 or Raid 10. I assume you did not do that due to space concerns or the database doesn't run at a high load factor.

If this is production, get some kind of a backup fast.

SEP