SCSI error 'opcode=0x5f' caused DMP to disable paths during shared diskgroup import

book

Article ID: 100027354

calendar_today

Updated On:

Resolution

Enabling  DMP debug logging

1.) Disable dmp_fast_recovery (on each node)
# vxdmpadm settune dmp_fast_recovery=off
 
2.) Increase dmp_debug_level to 6 (on each node)
# vxdmpadm settune dmp_log_level=9
 
3.) Increase vxconfigd logging to 9
# vxdctl debug 9 /var/tmp/vxconfigd.log

Issue resolved after upgrading the Brocade HBA Firmware and Driver version from 1.0.0.4 to 3.0.3.0.

# fcinfo hba-port
...
Manufacturer: Brocade
Model: Brocade-815
Firmware Version: 3.0.3.0
FCode/BIOS Version: FCHBA1.1.0.10
Serial Number: XXXX
Driver Name: bfa
Driver Version: 3.0.3.0 12/07/2011 20.25.14
Type: N-port
State: online
Supported Speeds: 1Gb 2Gb 4Gb 8Gb
Current Speed: 4Gb
...

 

Applies To

SFCFS 5.1SP1RP2 for Solaris x86
SunOS 5.10 Generic_144489-17 i86pc i386
Sun Fire X4640

Brocade HBA Driver v1-0-0-4

   PKGINST:  bfa
      NAME:  Brocade Fibre Channel Adapter Driver
  CATEGORY:  system
      ARCH:  sparc&i386
   VERSION:  v1-0-0-4
   BASEDIR:  /
    VENDOR:  Brocade
      DESC:  32 bit & 64 bit Device driver for Brocade Fibre Channel adapters
    PSTAMP:  20080115150824
 

Issue/Introduction

The cvm service group comes online on first node and the shared diskgroup gets import but it looks like the SCSI3 PGR registration receives error on one or both the paths. This causes DMP to disable one or both the paths. After enabling DMP debug logging, the messages logged to messages logs shows that DMP receives failure from the OS/HBA – EIO (5) error. This error seems to be returned by HBA Driver/OS and not from the Array. Error Message