topic Re: errors from the I/O subsystem in Operating System - HP-UX

errors from the I/O subsystem

James Hamilton_6 — Fri, 12 Mar 2004 10:23:25 GMT

Hi,

I'm running an oracle database on my HP9000 HP-UX 10.20 server. Oracle threw errors reporting a corrupted database file. I then looked in the syslog and found:

DIAGNOSTIC SYSTEM WARNING:
The diagnostic logging facility has started receiving excessive errors from the I/O subsystem. I/O error entries will be lost
until the cause of the excessive I/O logging is corrected. If the DEMLOG daemon is not active, use the DIAGSYSTEM command in SYSDIAG to start it. If the DEMLOG daemon is active, use the LOGTOOL utility in SYSDIAG to determine which I/O subsystem is logging excessive errors.

There are no other errors in the syslog. From reading other posts here, the signs point to a dodgy disk. Does this error automatically mean the disk is dead? Is my only option to bin that disk?

The only way I have of identifying which disk is reporting errors is from where the corrupt oracle datafile resides. Is there some way to extract more useful information?

If the disk is no good how do I go about replacing it? Can I just back it up to tape then restore onto a new disc. Will the corrupt data blocks cause problems on the new disk?

Any input gratefully received.
James.

Re: errors from the I/O subsystem

RAC_1 — Fri, 12 Mar 2004 10:35:10 GMT

Does syslog.log gives any further message?
Check /var/opt/resmon/log/event.log. This should give you some insight on what disk exactly is the problem.

How many disks you have? You can do pvdisplay on them and check. Also make sure if all are claimed in ioscan.

Anil

Re: errors from the I/O subsystem

Kent Ostby — Fri, 12 Mar 2004 10:50:01 GMT

This generally means that something (usually memory or an I/O) is generating a lot of errors.

If you have a support contract with HP, they can walk you through looking at the diagnostic logs.

If you have an idea which disk is bad, you could use dd to do a rough check.

dd if=/dev/rdsk/c0t6d0 of=/dev/null bs=24k

If it comes back with I/O error then its bad.

If it doesnt come back with I/O error then it still could be bad and you'd have to fall back on other means to diagnose it.

Best regards,

Kent M. Ostby