On Thu, 9 Sep 2004, ING. A. Neij wrote:

> I wonder how to define VIPA correctly.
> After defining VIPA for the TCPIP guest I can reach/ping  the z/VM LPAR at
> three IP addresses (see under for extract of profile tcpip).

Your profile looks okay, with one observation:

> HOME
>   10.12.100.10 VIPA1  ; <-- SOURCEVIPA for ETH0 and ETH1
>   10.12.100.08 ETH0
>   10.12.100.09 ETH1
> ; (End HOME Address information)

This obviously works for you, but might lead to problems due to ARP
caching in your hosts and routers.  VIPA addresses are normally allocated
in a separate subnet from the physical network interfaces, so that the
VIPA address itself is never directly associated with an ARP entry.  That
way if a network link fails, routing will take care of finding a new path
to the destination and you don't have to worry about old entries in ARP
caches.

> When I take a look at the devices I can see OSA device 1500 is used but
> OSA device 1300 is not used???

I'm going to assume that you want traffic to be load-shared across the two
ports...  This will not happen automatically.

The route table in a TCP/IP host controls which interface is used for
outgoing traffic.  Basically, the IP stack will look for the most specific
route that matches the destination of the packet to be transmitted.  When
the route table shows more than one entry for a given route (ETH0 and ETH1
in your case), even if that's the default route, it will simply use the
one that's higher in the table -- in your case, it seems that is ETH0.

Remember that VIPA is first-and-foremost a redundancy enabler for higher
availability.  It does not address load sharing or balancing.

> Where else do I have to define other/additional  OSA devices e.g. to use
> them with VIPA?

If you're at z/VM 4.3 or higher, check out your TCP/IP manual for
EQUALCOSTMULTIPATH in the ASSORTEDPARMS.  This allows traffic to be
balanced over equal cost routes (i.e. adding a little more smarts to the
decision path I described above).  This works for static routes as well as
OSPF or RIP, but if you're using ROUTED for dynamic routing it will get
turned off (so migrate to MPROUTE :).

Note that this will only alter the behaviour of traffic that leaves your
TCP/IP stack.  It does not change how traffic is sent to you.  If you
stick with your current config you may end up with a fairly even
distribution anyway (due to some randomness in the timing of ARP responses
from your OSAs), but that's not to say that it's a reason to stay that way
;)

Cheers, hope this helps,
Vic Cross

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