The IBM z/VM v5R1.0 manual describes in chapter "2.2.3.7 Virtual Switch
Failover".
In earlier versions of z/VM, the Tunk Port failure was detected as documented,
however in z/VM 5.1 an OSA 'Hang Detection' failover was introduced as follows
2.2.3.7.1 OSA Express Hang Detection
z/VM uses a timestamp and a full buffer condition to recognize when a real OSA
Express device associated with a Virtual Switch is stalled, or a
Guest LAN, which has defined a NIC for a simulated OSA Express, is stalled.
This hang detection works in two ways:
1. When CP determines that the buffers associated with the OSA Express device
are full and a timeout value passes with no relief, an error
message is issued and an attempt is made to fail the Virtual Switch over to
a backup device. This causes the TCP/IP stack controller to start
cleanup of the device, and error message HCP2832E is issued.
2. When CP determines that the buffers associated with the simulated OSA
Express device are full and a timeout value passes with no relief, it
assumes that the Guest LAN is hung and the device is disabled with an error.
An IBM Redpaper written by Vic Cross:
http://www.redbooks.ibm.com/redpapers/pdfs/redp3657.pdf details this redundancy
setup, however it talks only of z/VM 4.4 and thus does not contain any
information on the new 'Hang Detection'
We have successfully testing the Trunk Port failure by effectively taking one
port ofline, the VSWITCH then switches to the backup OSA port.
Thus far we have been unable to successfully verify the 'Hang Detection' nor
accurately say under what precise conditions it will be invoked.
Does anyone have any experience of this VSWITCH/dual-OSA-port setup and how to
go about testing the new "hang detection"?
Thanks in anticipation,
Dean Chapman
----------------------------------------------------------------------
For LINUX-390 subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit
http://www.marist.edu/htbin/wlvindex?LINUX-390
