Oracle ASM Fails With EIO 5 (SVR4 Error: 5: I/O error) When Removing One Path Of N From Mpxio With logical-block Balance Mode - Caused The Database To Crash
Last updated on APRIL 07, 2017
Applies to:Sun StorEdge 9900 Hardware - Version All Versions and later
Solaris Operating System - Version 10 3/05 and later
Information in this document applies to any platform.
3 x 25K Solaris 10 domains running an Oracle RAC cluster. The database is running on the top of some LUNs coming from an HDS 9990 array.
These LUNs have four paths available.
A single path to the storage is removed. To do so, unmap the LUN on Storage Navigator for one CHA port, by removing them from the associated host group.
When this is done, MPxIO detect the corresponding path going offline, but ASM software ( from Oracle ), place the drives to offline status.
This is recovered by mapping the LUN again to the host group.
If mpxio load balancing is set to round-robin, this cannot be reproduced.
Example host server1, it has 2 dual SG-XPCI2FC-QF4 adapters installed:
From ASM Oracle engineer:
Sometimes ASM will not immediately report/detect the I/O issue at the same minute/second as the OS,
since ASM caches the ASM metadata (associated with the ASM disks) in memory,
therefore until ASM performs an I/O operation on the affected disk(s) the problem will be detected and reported.
SVR4 Error: 5: I/O error indicates an I/O problem accessing either the physical disks associated to the ASM disks and/or the physical disks associated to the filesystem which contains the ASM Oracle Home.
ASM is working as expected , as there is no redundancy on the diskgroup.
ASM configuration is EXTERNAL REDUNDANCY , does not do mirroring.
it relies on hardware RAID redundancy in the 9990 storage.
Sign In with your My Oracle Support account
Don't have a My Oracle Support account? Click to get started
Million Knowledge Articles and hundreds of Community platforms