Article Reference
Article ID: 9001
Last update: 09-18-06 Issue:
I am seeing many SCSI errors with return code 0x20000 in my logs. This is causing DM/MPIO to fail and can even crash the system. What does this mean and how do I fix it?Resolution:Release Found:Red Hat Enterprise Linux 4 Update 2 or later
This is an example of the error code referenced in this article:
SCSI error : <3 0 0 6> return code = 0x20000
end_request: I/O error, dev sdg, sector 2879This error is caused by a reset on the fiber channel switch. This causes the multipath agent to become confused. There is a simple kernel module option that can be used to stop the multipath agent from becoming confused. To add this option edit the /etc/modprobe.conf and add the following line:
options qla2xxx ql2xprocessrscn=1Note: If a line that begins with options qla2xxx already exists in the modprobe.conf file, then simply append ql2xprocessrscn=1 to the end of the line
Once the file has been saved, regenerate the initial ram disk image file:
mkinitrd -f /boot/initrd-`uname -r`.img `uname -r`After the initrd is rebuilt, reboot the system for the changes to take effect. The Qlogic HBA card should not throw errors after a reset from the switch. This causes DM MPIO to remain stable and has no negative effect on perfomance.