So i read that on lag connections you are supposed to disable spanning tree. i'm guessing that is port level and not lag level correct? it makes sense.
I'm not aware if that is best practice or not. I've been leaving it enabled on my phy port agg memebers to guard against operator error of mistakenly removing port from agg while the other side is a single logical port.
I'll try not to rattle on, and hopefully, this will actually make sense.
Looks like you were asking about RSTP States, both globally and at the port level.
I have some rules I found helpful.
Basic rules of thumb:
NO EMT NNI ports should be in the FWD state while it is a member of an AGG.
Only the virtual AGG port will be in the spanning tree FWD state.
Management VLAN 4092 (or 4091, 127) can only be FWD’ing on one port at a time.
The reason behind this is that upstream STP NNI port will take care of the STP rules, and block or forward.
This is to lift the Ciena device from the burden of loop prevention, and to control the FWD'ing of all ports across the Ciena (and attached UNI ports).
For instance, The 7450 upstream from the (5150) ciena device has RSTP globally enabled, and it is Root. The (virtual) AGG has RSTP enabled on the virtual port(s).
The Ciena (ex: 5150) has RSTP globally disabled, but Enabled on the AGG virtual port. The downstream facing UNI ports have RSTP enabled on their ports,
and the attached/cascaded device (another ciena 393x?) has RSTP disabled globally and on their phy ports to the UNI upsteam 5150 device.
While an AGG is being configured, RSTP should be globally disabled.
With that said, Here is some logic behind all this STP stuff.
The rules for Spanning-tree in regards to a virtual LAG port, as I understand them:
Configuration: 4-port AGG, a PRI VLAN, not SEC MGMT vlan is being used. VLAN-1 is not part of any phy port.
!) The Admin user is configuring a multi-port AGG2) the Primary port of the LAG is the lowest numbered port by default and design (ie vlan 4092)
3) the PRI port is the only physical port in the AGG on the PRI Management VLAN - which is normally NOT ever VLAN 1
4) the remaining 2-3 ports of the AGG are not members of the PRI Management VLAN
5) if a secondary (redundant) OOB Management Vlan is available, it is assigned to the 2nd phy port of the AGG (ie vlan 4091)
6) the remaining 2nd and 3rd PHY orts of the AGG should be assigned to VLAN 1,127, but usually only in VLAN 1
Initially (in the 'base' startup config file), RSTP is enabled globally. All the physical ports and the LAG virtual port will have RSTP enabled by default, and RSTP is placed into the RSTP Mode (not MSTP Mode).
MSTP is disabled globally.
A 5150 RTSP SHOW command screen shot:
Are you able to provide any insight on how STP on a ciena switch interacts with other vendors? I've always had the understanding that stp is a standard and it doesn't matter who is talking to who. However, while getting ready to make some network changes, a coworker has told me that ciena has it's own implementation of stp and connecting aggs to dell switches is a bad idea. Is there any merit behind this?
A very simplified concept of what I'm trying to do is this: There's a 5305 switch, and 2 dell M6348's in a blade chassis. Those 2 dells are currently connected together with an agg between them and both have aggs of their own going to a third dell switch. Spanning tree works fine in this setup.
We're trying to clean up and I want to take both of those aggs directly to the 5305. Anyone see any issue with doing that? As long as stp is configured properly like is mentioned above?
wow! the 5305 is an older beast. So I'm a bit hazy on the STP implementation on it. But in General for the Ciean switches, 5150/393x/5140, Ciena wants/lets the NNI upstream device handle the STP looping mechanics. The reason is to prevent UNI users from falling into a loop on their net without interference from the other VLANs running on the box. Its not only cleaner, b ut saves bandwidth and memory consumption on the Ciena, causing a total box lock up or failure. Usually th upstream divices have a much better CPU for this. Besides, Ciena as a Root is not where these boxes operate normally.
What I mentioned above about the AGGs earlier still applies as far as the virtual port taking over for the PHY ports in the AGG itself.
I have not heard of any issues with Dell on STP. I would need more specifics to comment further. Ciena is Fully standard compliant though. they do however, have an approach on STP when it comes to RSTP vs MSTP Modes running on the Switches. This again, was done to conform to various forms of STP protocols, and not some hack job.
As with all AGGs, provisioning is the key. MTU and ether-type must match, as well as any other parameters on both ends in the link.
If you are having issues with this working properly, or are confused as to what is the issue, please post the Question separately in this forum, with the configs and a basic topo diagram for us.
Commands of interest in SAOS, not necessarily the 5305, are: mstp shoiw, rstp show, agg show <member>, agg show agg, po sh status, port show eth, port show config; to name some good commands.
Retrieving data ...