Stacking Procurve Switches (nested)

[Simon Whiting]

G'day,

We have five HP 4108gl switches as our set of core switches. Currently we have them in a cascaded arrangement, with a two port trunk per cascade connection, eg

SW-1 to SW-2 (2 port trunk)
SW-2 to SW-3 (2 port trunk)
SW-3 to SW-4 (2 port trunk)
SW-4 to SW-5 (2 port trunk)

This configuration has served us extremely well except on the rare occasion when one of the middle switches has failed, resulting in a split network.

Due to the length of the stacking port cables (1/2 metre) it is impossible for us to connect SW-1 to SW-5 in a cascade loop. We propose to break the trunked stacking ports and configure the switches in a staggered switch format, ie

SW-1 to SW-2 and to SW-3
SW-2 to SW-3 and to SW-4
SW-3 to SW-4 and to SW-5
SW-4 to SW-5

Is this configuration likely to cause any problems?

All suggestions/recommendations greatly appreciated.

Cheers,
Simon Whiting.

[Jeff Brownell]

this looks fine. have you had success in testing?

[stephane_12]

Another idea is to put an (or some) Ethernet cable between the 1 and the 5 and activate the Spanning Tree Protocol ..

No ?

Stephane

[Preston Gallwas]

on Anoter note, isnt less than 1m cable hurtful for some standard or another, with regard to crosstalk and the like? I Could be loony...

[Simon Whiting]

We are currently looking at using spanning tree on three interconnected test switches before touching our core switches.

We are trying to optimise the setup by only enabling spanning tree on the ports interconnecting these three switches. The hope is to limit spanning tree packets to only interconnection ports and to enable STP fast mode on all other.

The three port types we have identified were:
1). switch interconnects for core switches with redundant paths
2). switch interconnects for leaf switches without redundant paths
3). directly connected device (eg servers)

Only port type "1" needs spanning tree as only core switches will have redundant paths. Port types "2" and "3" will be single connections and not require spanning tree.

On our three interconnnected test switches we are currently testing STP configuration settings such as "Edge", "Point-to-Point" & "Mcheck", and have disabled "cdp" and "lacp" on all ports to try and achieve the above. Unfortunately we are still seeing STP broadcast packets on end nodes/devices, however the fast mode seems to behaving as expected.

Any further suggestions?

[Nick Hancock]

Simon,

All of the options you have used still assume that Spanning Tree / RSTP is in operation therefore you will still see BPDUs. I believe the only way you are going to stop the BPDUs on server ports is by actually disabling STP on those ports ( not even sure whether this is possible on a per port basis on this switch) - generally inadvisable anyway because it just adds risk of loops being created by accident. Fast-start does a perfectly acceptable job.

The only caveat that I have seen on this recently is PCs with bridging mode turned on. This makes the PC/Server send out legacy Spanning Tree BPDUs which can mess up the fast-start.

Cheers,

Nick

[Simon Whiting]

G'day All,

Thanks for all the replies. With have now enable RSTP. We have testing various configurations on a set of test 4108GLs by tweaking vaious RSTP port settings and found that the default RSTP settings work best with LACP disabled and CDP enabled.

Thanks again to all that replied.

Cheers,
Simon.