Sense and nonsense of Stacking ProCurves

[Marc Haber]

Hi,

I have recently spent some time with the Stacking features of the ProCurve Switches (especially the 2848) and am now wondering which advantages stacking is supposed to give.

Here is what I found out
(a)
A new switch can be pulled into a stack by simpl connecting it to the same broadcast domain as the stack controller and giving the "stack member mac ". This eliminates the need to work with the serial console during initial configuration, but this can also be achieved with a DHCP server and telnet - the switches are DHCP clients in the default configuration and allow passwordless telnet.

(b)
All switches automatically have the same password and central SNMP communities. I actually consider this a misfeature, since I usually use switch-specific passwords and the central SNMP communities do not override the default "public" community.

(c)
I can use "telent " on the commander to telnet to an arbitrary stack member without password. This is actually a loss of security since I usually use ssh with keys which has the same look-and-feel, but with encryption and authentification.


And this is the list of things that I would like to have in a stack, but it looks like these are not available:

(1)
Possibility to configure ports on different switches within the same command:
|conf t
|vlan 108
|untagged 1:1,2:4,5:2-5
to configure switch 1 port 1, switch 2 port 4 and switch 5 port 2-5 in VLAN 108, untagged, or
|conf t
|int 1:1-4,2:10-18
|duplex full
|speed 100
to set switch 1 port 1-4 und switch 2 port 10-18 to 100Full

(2)
The possibility to see all configuration with a single "show run" and to be able to change the configuration of all switches with a single "conf t"

(3)
A switch-spanning MAC address table which would enable me to directly know switch _and_ port a given MAC address is connected to with a single "show mac-address" command on the switch commander.


Did I miss something, or does ProCurve Stacking really not offer any of these possibilities?

Greetings
Marc

[Matt Hobbs]

It's pretty basic. I think the disadvantages of it outweigh any advantages. I would only use it if I absolutely had ran out of IP addresses.

[André Beck]

Hi Marc,

IMO this "stacking" is a me-too-implementation of what Cisco called switch clustering, a pure management-wise coupling of several switches that otherwise stay separated bridges. As Matt already said, the only true gain is in needing less IP addresses for management. For some reason, this appeals to a certain class of customers (usually those without any sound network design - else there wouldn't be a VLAN to span more than one access layer switch).

The bad thing is that they burn the word "stacking" which IMO is something entirely different. True stacking is what you describe: Single configuration, single bridge identity, single router identity (on L3 switches), single hop, multi-member aggregates. In other words, a chassis collapsed into some cables and without dedicated supervisors (every stack member has to bring in enough CPU for that). So far, the only true contemporary stack I know of is Cisco StackWise (3750(E)). But of course I don't know every vendor's product lines.

HTH,
Andre.

[RicN]

If you take a default-configured switch, is there a risk that the switch might be "taken over" by some random commander-switch available on the network?

I have seen some recommendation to run "no stack" to aviod this, but is it correct that a default switch could join some stack without any configuration at all?