On Wednesday, 06/02/2010 at 03:18 EDT, Offer Baruch
<[email protected]> wrote:
> Just to be clear... I think we are talking about different things...
when
> defining a port at the switch level you can define 1 or more VLANs in
trunk
> mode and 0-1 VLAN in access mode. I called the access mode VLAN as the
> native VLAN...

Yes, terminology is extremely important.
o  There are *trunk* ports and *access* ports (Cisco terminology)
o  Access ports accept and emit only untagged frames.  Every access port
has an associated *port VLAN ID* (PVID) which is logically added to
inbound frames and stripped from outbound frames.  The default PVID is
usually VLAN 1.  So even if you don't explicitly assign a PVID, there is
always exactly one VLAN ID associated with an access port.
o  Trunk ports accept and emit both tagged and untagged frames.  Untagged
traffic on a trunk port is associated with the *native* VLAN ID (default
native VLAN ID is usually VLAN 1).
o  All trunked switches must have the same native VLAN ID.  Hence, you
must use the NATIVE keyword on DEFINE VSWITCH if the native VLAN id of the
attached switch is not VLAN 1.

When the native VLAN ID and all port PVIDs are 1 (the usual default
configuration), you have what looks like a hub.  All untagged traffic on
any port, trunk or access, is in VLAN 1.  There are some in the industry
who think that is a Bad Thing, but it is what it is.

> z/OS was working in access mode (not VLAN 1) before z/VM came into the
> picture...

Terminology: Software is either "VLAN aware" or "VLAN unaware".

> when z/VM arrived new VLANs were added to the port in trunk
> mode... VSWITCHs are only working with VLAN tagging... no host is using
VLAN
> 1 (as defined in all of our switches).
> Do you still think an accident waiting to happen? If so please elaborate
> some more...

Yes.  If z/OS is VLAN unaware, then z/OS is emitting untagged frames. That
means they will be associated with the switch's native VLAN ID.  I think
that is a Bad Thing.  SOME of the switch management protocols still
operate on the *untagged* frames.  IMO, you need to add a VLAN
specification to z/OS and change the native VLAN ID back to 1 if you're
going to share the OSA that way.

People often ask me if they can share an OSA between a VSWITCH and
something else.  I say "You *can*, but you *shouldn't*."  This is
precisely why.  z/OS and other VLAN-unaware entities can share an OSA.
VLAN-aware entities *should not* unless they are simply replicants.  That
is, LPARs 1 and 2 both run z/VM with a VSWITCH called "INTRANET".  It is
ok to share an OSA because you have simply split the workload of a single
INTRANET connection (VSWITCH) among two LPARs.  Of course, you can't use
Link Aggregation.

But, IMO, HOST connections should not be mixed with VSWITCH connections.
Hosts are not switches and there is a reason hosts are not plugged into
trunk ports.  Different rules, different expertise and training, different
expectations, a different price to pay for misconfiguration.

As I said, it's not about what you *can* do, it's about what you *should*
(or should not) do.

Alan Altmark
z/VM Development
IBM Endicott

----------------------------------------------------------------------
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 more information on Linux on System z, visit
http://wiki.linuxvm.org/

Reply via email to