topic Re: SCSI timeout on raw devices never happens in Operating System - HP-UX

SCSI timeout on raw devices never happens

Augustas Gutautas — Mon, 19 Mar 2007 11:38:04 GMT

Hello,

we are testing Oracle RAC with ASM configured with Normal redundancy with two EMC Symmetrix DMX-3 storage arrays. All Oracle database's datafiles are placed on ASM disks which are separated into two failure groups. When one of the storage arrays are turned off (for testing) the Oracle database seems to 'hang' - it produces no errors but queries 'hang' and wait for output from scsi disks which are offline.
The problem probably lies in HP-UX scsi configuration which does not escalate an I/O error to upper layers when offlined scsi disks are accessed. It queues infinitely all the requests. When the offlined storage is up again the Oracle database operates as usual and all the 'hung' queries continue and produce an expected output.
How the scsi parameters of HP-UX may be changed to report an I/O error to upper layers?

HP-UX version 11.23 (Itanium)

Augustas

Re: SCSI timeout on raw devices never happens

Court Campbell — Mon, 19 Mar 2007 12:15:42 GMT

Can you explain what you mean by Normal redundancy?

Re: SCSI timeout on raw devices never happens

Augustas Gutautas — Mon, 19 Mar 2007 12:28:44 GMT

Hi,

Normal redundancy in Oracle ASM means that Oracle software makes a mirror of data between two storages.

Augustas

Re: SCSI timeout on raw devices never happens

Eric Antunes — Tue, 20 Mar 2007 10:04:14 GMT

Hi Augustas,

Check Metalink Note 419014.1 and 303760.1

Best Regards,

Eric Antunes