[f-nsp] Spanning-tree event on single VLAN brings down LAG?

Thu Oct 11 18:33:08 EDT 2018

We did not expect the “1/3/8, 2/3/8, 3/3/4, 3/3/8” LAG to drop, either – two network engineering consultants that I’ve shared this with don’t have good theories yet, either.

You’re right, the ICX6610 could have just been documenting the LAG member drops instead of initiating it, I just don’t know how to figure that out.

Frank 

From: Howard, Christopher <Christopher-Howard at utc.edu> 
Sent: Thursday, October 11, 2018 8:55 AM
To: foundry-nsp at puck.nether.net; frnkblk at iname.com
Subject: Re: [f-nsp] Spanning-tree event on single VLAN brings down LAG?

Looking at your log, 1/3/6 is where the block event happened. I would not expect the 6610 to have dropped the lag because of this, no matter which version of spanning tree you're running, because 1/3/6 is not a part of that lag.  

Just some ideas: 

- It could have been dropped on the other end by either the same or an unrelated event and you're just seeing the documentation of that on the 6610.  

- I've seen before where a loop drove the switch CPU high enough to flap lags. Since spanning tree blocked the port that should have protected the CPU, unless there are other ports not participating in spanning tree on that vlan.

- A coincidence, but I admit the chances of that are unlikely with the timing of the lag drop.

-Christopher

On Thu, 2018-10-11 at 08:07 -0500, frnkblk at iname.com <mailto:frnkblk at iname.com>  wrote:

Christopher,

Thanks for sharing.  Your explanation regarding vlan enablement versus port disablement makes sense to me.  

I plan to turn up 802.1w in a future maintenance window.

>From the Brocade ICX6610 logs is it clear if the ICX6610 turned down the 10G ports, or did the core router link it down and the ICX6610 was merely documenting that it was being pulled down?

Frank 

From: Howard, Christopher <Christopher-Howard at utc.edu <mailto:Christopher-Howard at utc.edu> > 
Sent: Wednesday, October 10, 2018 12:32 PM
To: foundry-nsp at puck.nether.net <mailto:foundry-nsp at puck.nether.net> ; frnkblk at iname.com <mailto:frnkblk at iname.com> 
Subject: Re: [f-nsp] Spanning-tree event on single VLAN brings down LAG?

One vlan getting blocked by spanning tree should not bring down the lag. The vlan should only block on the interfaces required to remove the loop. As far as I know, if you have spanning tree enabled on the vlan, but disabled on the physical port, then the physical port no longer participates in spanning tree for any vlan.

Here's config from one of my switches. This is an ICX7750 with 8.0.30r code, same code that ICX6610 can run.

lag lag-name dynamic id 1
 ports ethernet 1/1/2 ethernet 2/1/2 
 primary-port 1/1/2
 deploy
!
vlan 123 name vlan-name by port
 tagged ethe 1/1/2 ethe 2/1/2
 spanning-tree 802-1w
!
interface ethernet 1/1/2
 spanning-tree 802-1w admin-pt2pt-mac

A snipped output from "show 802-1w" below with vlans in both discarding and forwarding states on that lag.

--- VLAN 404 [ STP Instance owned by VLAN 404 ] ----------------------------

1/1/2     128 2000     T   F    ALTERNATE  DISCARDING  2000      0001609c9fd73700 

--- VLAN 999 [ STP Instance owned by VLAN 999 ] ----------------------------

1/1/2     128 2000     T   F    ALTERNATE  DISCARDING  2000      0001609c9fd73700 

--- VLAN 2007 [ STP Instance owned by VLAN 2007 ] ---------------------------

1/1/2     128 2000     T   F    ROOT       FORWARDING  0         0000609c9fd73700 

--- VLAN 2057 [ STP Instance owned by VLAN 2057 ] ---------------------------

1/1/2     128 2000     T   F    ROOT       FORWARDING  0         0000609c9fd73700 

-Christopher

On Wed, 2018-10-10 at 11:59 -0500, Frank Bulk wrote:

We had network event deeper in our network that resulted in some kind of
spanning tree event as evidenced by a TC (topology change) on our L2-only
ICX6610 stack.  What surprised us was that one VLAN going into a blocking
state resulted in the south bound LAG going down.

This is probably not right, but the root bridge for VLAN 294 is on the
ICX6610 stack. VLAN 294 is on both LAGs (and no other port/LAG)

Three questions:
a) when a STP event occurs is it normal that a VLAN goes into Blocking on
all the ports where it is present?
b) when a VLAN goes into blocking should all the physical (or LAG) ports
associated with that VLAN go down?
c) does configuring spanning tree on the port-based VLAN but disabling
spanning-tree on the physical (or LAG) ports prevent the physical port (or
LAG) going down?

Frank

============================================================================
====
Oct  9 21:35:22 STP: VLAN 294 Port 1/3/6 Bridge TC Event (MakeBlking) 
Oct  9 21:35:22 STP: VLAN 294 Port 1/3/6 STP State -> BLOCKING (MakeBlking) 
Oct  9 21:35:25 System: Logical link on dynamic lag interface ethernet 2/3/8
is down. 
Oct  9 21:35:25 System: Logical link on dynamic lag interface ethernet 3/3/4
is down. 
Oct  9 21:35:26 System: Logical link on dynamic lag interface ethernet 3/3/8
is down. 
Oct  9 21:35:29 Trunk: Group (1/3/8, 2/3/8, 3/3/4, 3/3/8) removed by 802.3ad
link-aggregation module. 
Oct  9 21:35:29 STP: VLAN 4 Port 1/3/8 STP State -> BLOCKING (PortDown) 
Oct  9 21:35:29 STP: VLAN 4 Port 1/3/8 STP State -> DISABLED (PortDown)
etc.

 Event
   |
Transport
  | | 
  | | (1/3/6 & 2/3/6)
  | | 
ICX6610 (L2)
 | | | |
 | | | | (1/3/8, 2/3/8, 3/3/4, 3/3/8)
 | | | |
Core (L3)

_______________________________________________
foundry-nsp mailing list
foundry-nsp at puck.nether.net <mailto:foundry-nsp at puck.nether.net> 
http://puck.nether.net/mailman/listinfo/foundry-nsp

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://puck.nether.net/pipermail/foundry-nsp/attachments/20181011/b443d160/attachment.html>