This operational problem with passive backup configurations just never
goes away.  After over 30 years of networking I consider this design to
be a bad practice nearly guaranteed to fail.

If you want the backup path to be there when you need it, then it must
be in production, period.  An active backup path, meaning with a live
unique IP address, allows for production monitoring.  I further advocate
that all paths support dynamic routing protocols to provide automatic
recovery with applications utilizing a loopback address advertised to
the network for connectivity.  Most routing protocols will recover in a
time frame that is short enough to allow active TCP/IP connections to
stay alive during the failover, making the failover transparent to the
applications and eliminating end user outages (just a bit of a slow down
while failover occurs).  The quickest is OSPF.  All mainframe routing
protocols provide mechanisms to eliminate the entire corporate routing
table from being managed by the host instance, just the default route.
This design further can allow for not just an active backup path but
also load sharing between the paths.

In other words, "VIPA," adapted to the specific OS' networking
capabilities, is a best practice.

This post is not meant as a criticism of Mr. Levy's situation which may
have constraints, technical or organizational, that preclude these best
practices from being adopted, but as a general statement of networking
best practice for high availability.  Both the host
administrators/systems programmers and router administrators must
acknowledge that the network does not stop or begin at the Ethernet
cable and work together.  Networking has its equivalent of "gen jockeys"
who can configure but do not truly understand the technologies with
which they work.

Harold Grovesteen

Levy, Alan wrote:



We have a vswitch defined here as follows:



VSWITCH SYSTEM VSW1     Type: VSWITCH Connected: 3    Maxconn: INFINITE


 PERSISTENT  RESTRICTED    NONROUTER                 Accounting: OFF


 VLAN Unaware


 State: Ready


 IPTimeout: 5         QueueStorage: 8


 Portname: UNASSIGNED RDEV: FA00 Controller: DTCVSW2  VDEV:  FA00


 Portname: UNASSIGNED RDEV: FB00 Controller: DTCVSW1  VDEV:  FB00
BACKUP



This vswitch was defined over 1 year ago and at that time failover was
tested successfully.



This works great and worked great until this past Sunday. Our network
group was doing maintenance and brought down the primary osa FA00. The
switch did not fail over to the FB00 osa. They had to back out their
changes.



We found out this week that sometime in the past year, one of the cables
to the FB osa became unplugged (there are no other servers connected to
that switch).



Our network group actively checks the LAYER-2 (MAC) addresses display on
the Cisco switches where the OSA connections are plugged.  They only see
the MAC address of the primary interface on the VSWITCH.  The backup
VSWITCH port does not show a MAC address on the Cisco switch.  They
would like to actively monitor both interfaces so that if a real problem
occurs on the backup interface, it can be detected PRIOR to a VSWITCH
failover.  Since they currently can't monitor those connections they
never noticed it had a problem.  Then when we really needed the backup
it wasn't there and we experienced an outage.



Is there a way to have both the primary and backup OSA interfaces show a
MAC address on the Cisco switches?








----------------------------------------------------------------------
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




----------------------------------------------------------------------
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

Reply via email to