IBM Support

Oracle Databases with ASM and Asynchronous I/O Enabled on HPUX Hosts May Timeout if a Node or Node Canister is Reset or Power Cycled

Flash (Alert)


Abstract

If a Storwize V7000 node canister or SAN Volume Controller node is power cycled or reset when using Oracle ASM with asynchronous I/O enabled on a HPUX host, the Oracle database may suffer a timeout resulting in a loss of access.

Content

The following two HPUX parameters should be changed from the default 30 seconds to a new value of 120 seconds with /dev/async enabled:

  • ESD_SECS=120

    This attribute determines the timeout of I/O operations to block level devices.
  • asyncdsk_io_timout=120

    This attribute determines the overall timeout for asynchronous I/O to block devices.

The above parameter changes will allow sufficient time for Oracle to sustain access during a node or node canister reset or power off scenario.

Alternatively, this issue can also be resolved by disabling asynchronous I/O.

Cross reference information
Segment Product Component Platform Version Edition
Disk Storage Systems IBM Storwize V7000 (2076) 6.4 IBM Storwize V7000 6.3, 6.4
Storage Virtualization SAN Volume Controller 6.4 SAN Volume Controller 6.3, 6.4

Document information

More support for: SAN Volume Controller
6.4

Version: 6.4

Operating system(s): Platform Independent

Reference #: S1004074

Modified date: 14 June 2012