Disk Errors during boot and run time.

[Bryn M. Reeves]

On Fri, 2009-03-06 at 12:22 +1300, Paul Ward wrote:
> # ls /boot
> ls: reading directory /boot: Input/output error

What's in dmesg at this time?

> I have been told that the disks use multipath but I have no experience
> of this to date.
> I know the disks are on a SAN but as yet have not been able to locate
> them using the IBM SAN manager.
> 

> Linux version 2.6.18-53.1.21.el5PAE

So, RHEL5.1?

> (brewbuilder at ls20-bc2-13.build.redhat.com) (gcc version 4.1.2 20070626
> (Red Hat 4.1.2-14)) #1 SMP Wed May 7 08:56:33 EDT 2008

>   Vendor: IBM       Model: 1814      FAStT   Rev: 0916
>   Type:   Direct-Access                      ANSI SCSI revision: 05

So it's an IBM FAStT SAN? These are active/passive storage arrays that
require use of a multipath hardware handler to properly manage switching
between the active and passive paths and preventing I/O being sent to a
controller that cannot handler it.

The I/O errors that you see are a result of things trying to access the
passive paths (e.g. partition scanning, lvm label scanning, udev/hal
probes etc.).

RHEL5.1 included the old device-mapper hardware handlers. These will
only take effect once multipath has configured the devices and only
handle path switching in the event of a path failure (i.e. you'll still
see I/O errors if something tries to access one of the underlying paths
directly rather than via the multipath device map).

RHEL5.3 introduces the scsi device handler framework as a replacement
for the device-mapper hardware handlers (this appeared upstream in
2.6.26).

Whether you decide to update or not it's probably worth carefully
checking the current multipath configuration on the system as this is a
very common area for configuration mistakes.

Regards,
Bryn.