10

I am trying to set up the following arrangement with 10 switches:

Switch arrangement

I think it's relatively self-explanatory, but basically I have 8 different networks (in 8 separate VLANs) which are physically switched separately too (just because I can).

I want to set up the switches for each "pair" of LANs, and for each endpoint switch have a backup path to the root switch via its mate.

I'm using HP V1910-48G (JE009A) for the ROOT and "ROOT B" switch.

  • This is where the servers are physically connected to the network.
  • The reason I have switches where the servers are located is to get access to 8 fibre ports.
  • There's a 4Gbit LACP trunk between ROOT and ROOT B to handle the bandwidth and all VLANs are trunked and tagged
  • All switches are managed on VLAN 1080. Does that matter? I'm thinking not.
  • All VLANs are tagged on the trunks as necessary. Again, I can't see it mattering.

For my endpoint switches, I'm using HP V1810-24G v2 (J9803A) switches.

All switches DO support RSTP!

At the moment I can't get this working correctly. When I disconnect the trunk between Switch with VLAN20 to Root switch - the port to Switch VLAN21 doesn't forward packets.

Bridge configuration for all Endpoint Switches: RSTP Bridge Configuration

Port config on VLAN20 Switch: VLAN20 Switch Port Config

Any help would be most appreciated.

Thanks, Sam.

  • When you say, "pull one trunk," do you mean you disconnect (for example) the link between the switch with VLAN 20 and the the "root"? Also is "root" the root bridge for all vlans? – Ron Trunk Sep 23 '14 at 12:01
  • 2
    You have explicitly set the priority on the ROOT, and checked that other switches see it as the designated root? – richardb Sep 23 '14 at 13:13
  • @Ron. Yes, I'm physically removing a cable. Yes, it's the root bridge for all VLANs. – user8179 Sep 23 '14 at 20:38
  • 1
    @Richardb: I set the priority to 0. I'll double check that all switches recognize it as root. Thanks. – user8179 Sep 23 '14 at 20:40
  • 1
    @Ron The 1810 v2 does. The 1810 v1 does not. I have v2. I'll get some screenshots happening and reply again soon. Thanks for continued help! – user8179 Sep 23 '14 at 23:56
2

Is HP's version of "RSTP" a Cisco-like Rapid-PVST+, is it more of 802.1q RSTP or something like 802.1s MST?

In the former case, you have to make sure that "Switch 21" - although all it's user ports are in VLAN21, also has VLAN20 configured, and is actively participating as a bridge in VLAN20's spanning tree (with only two STP interfaces, in that case).

If the config is more MST-like, you'll have to make sure that same mapping of VLAN-to-MST-Instance(s) is applied on the given (set of) switche(s). Plus, you have to make sure that the VLAN used to talk MST from switch to switch is available/allowed on all inter switch links (I believe this is usually done in VLAN 1 and without tags).

Essentially, this boils down to the same problem, no matter which dialect of spanning-tree you are running.

Both members of a such a switch pair need to be actively participating as bridges in the spanning-tree(s) for both VLANs of the given switch pair. Of course, the inter switch links must be configured to allow both VLANs; your diagram suggest that this is the case.

Cheers Marc

  • HP uses industry standard 802.1w RSTP. It's not compatible with Cisco R-PVST+ – Ron Trunk Sep 18 '17 at 1:22
2

Your design is valid, but you need to carry the management VLAN1080 as well to all switches. As your network isn't that huge, I'd just tag out every single VLAN to every switch (which is how Cisco's VTP works by default by the way) as you'll only see a limited amount of extra broadcast traffic that way.

The reason you can't get RSTP to work is that you have enabled almost every single feature of STP on every port.

Disable Root Guard, TNC Guard, BPDU Protect and especially BPDU filter on all ports, as these are meant to be used for very specific purposes (read up on what these do before you enable them on select ports). BPDU filter is likely the problem, as it basically tells the switch to not send BPDU (special packets that's used to determine the STP topology) or process the ones it recieves on that port, essentially disabling STP alltogether. It's basically a way to tell the switch to "what I'm going to do looks stupid, but run with it regardless.) You should also tweak the priority of your root switches to make sure that the spanning tree builds it's topology from the root switches instead of from some random switch with the lowest MAC address.

Please read through my guide on STP on HP switches, it's meant for the higher end HP Procurve models, but it goes through some core concepts of STP that are useful for you. https://grumpytechie.net/2015/02/27/how-to-set-up-stp-on-hp-switches-hp-networking-series-part-1/

0

You said:

When I disconnect the trunk between Switch with VLAN20 to Root switch - the port to Switch VLAN21 doesn't forward packets.

This is most likely due to how STP converges after a change in network topology. After the trunk from Vlan20 is shut down, the surrounding switches will change their STP behavior to accommodate the change. This can lead to ports being blocked which were not blocked before.

You should set one of your core switches to be the primary root for half of the vlans, and the other core switch to be primary root for the other half. This will typically allow all vlan traffic to reach all devices in a LAN.

-1

If it was me, I think I'd avoid the VLANs except on the core switches, and run two cables per edge switch. Edge switches are just access then, and maybe the simplicity would help.

enter image description here

  • That's not really a helpful answer, it doesn't help the OP figure out why he can't get RSTP to work, it doesn't help anyone that has a similar issue, and it's not really good design either (not using the full potential of the devices for example). – Stuggi Dec 20 '17 at 9:09
  • It was intended to suggest that perhaps a simpler architecture might nonetheless be useful, ie satisfying my guess of the asker's higher-level goal, as others have done here too. Using full potential of a given device very usually also introduces dependencies, which may incur complexities and unreliabilities later. Many ways to skin a cat, as they say. – jonathanjo Dec 20 '17 at 11:56
-2

Over-engineered. Those are SMB switches. You have approx 300 Ethernet ports - a small number. Keep It Simple - that leads to uptime and easy management.

Wipe configs clean, factory reset. Put a management IP address on each switch and set them up w basic monitoring.

enter image description here

  • any chance the downvoter(s) can identify what part of my answer is incorrect? Thanks in advance. – Ron Royston Sep 20 '17 at 3:56
  • I wasn't a voter, but your drawing doesn't seem to have any redundancy, and isn't the lowest VLAN number 1, not 0? – jonathanjo Oct 20 '17 at 18:40
  • @jonathanjo thanks for the feedback. I updated my answer deleting mention of any particular VLAN#. Right re redundancy. If a switch fails, they've got to replace it. Network layer redundancy could be achieved by enabling routing on each switch/uplink interface. RSTP, and STP in general, is overly complex especially in light of the low end platforms deployed. MLAG technologies were created for this very problem. – Ron Royston Oct 20 '17 at 20:26
  • I agree with you that complexity is the enemy, most especially fault-tolerant complexity if unmonitored! (Lots of people don't check their RAID, for example, and only notice after catastrophic failure). Re kit perhaps it's proof-of-concept work. – jonathanjo Oct 20 '17 at 21:03

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.