|
Posted by anoop on February 9, 2008, 10:47 am
If you were Registered and logged in, you could reply and use other advanced thread options
> I'm setting up a new small backbone network (and edges too, partly)
> and I figured I'd do it _right_ this time and create a bit of
> redundancy and stuff. Anyway, I'm a bit confused right now...
>
> But first - our setup:
>
> Switch A - HP ProCurve 6200yl (K.12.57)
> Switch B - HP ProCurve 6200yl (K.12.57)
> Switch C - HP ProCurve 2900 (
>
> Routing enabled in A and B. The 2900 in our main computer room.
>
> The idea is to configure them in a triangle - with primary links between
> C-A (4x1000Mbps) and B-C (4x1000Mbps) with a backup link A-B (1x1000Mbps)
> in case something goes wrong and to use MSTP to make sure the A-B backup
> link normally is "cut".
>
> The thing I'm a bit curious about is why STP only seems to block
> half (one direction) the A-B link (only one fiber link is used of
> the two we originally was thinking of using, the the second interface
> is administrative configured "down"):
This doesn't make sense. xSTP works on both directions of
a link. If a port is blocked, it should be blocking both incoming
and outgoing traffic, regardless of the version of STP
(STP/RSTP/MSTP). It might be a bug, or it might be due to
your configuration, but I don't see anything obvious that
was explain this behavior.
You could try posting your query at:
http://forums12.itrc.hp.com/service/forums/categoryhome.do?categoryId=269
> Router A # show span Trk2
> ...
> | Prio | Designated Hello
> Port Type | Cost rity State | Bridge Time PtP Edge
> ----- --------- + --------- ----- ---------- + ------------- ----- --- ----
> Trk2 | 20000 16 Forwarding | 001c2e-149f40 2 Yes No
>
> Router A # show int brief 3-4
>
> Status and Counters - Port Status
>
> | Intrusion MDI Flow Bcast
> Port Type | Alert Enabled Status Mode Mode Ctrl Limit
> ------- --------- + --------- ------- ------ ---------- ----- ----- ------
> 1-Trk2 1000SX | No Yes Up 1000FDx off 0
> 2-Trk2 | No No Down off 0
>
> Router B # show span Trk1
> ...
> | Prio | Designated Hello
> Port Type | Cost rity State | Bridge Time PtP Edge
> ----- --------- + --------- ----- ---------- + ------------- ----- --- ----
> Trk2 | 20000 16 Blocking | 001c2e-149f40 2 Yes No
>
> Router B # show int brief 1-2
>
> Status and Counters - Port Status
>
> | Intrusion MDI Flow Bcast
> Port Type | Alert Enabled Status Mode Mode Ctrl Limit
> -------- --------- + --------- ------- ------ ---------- ----- ----- ------
> 1-Trk2 1000SX | No Yes Up 1000FDx off 0
> 2-Trk2 1000SX | No No Down 1000FDx off 0
>
> The trunks are configured as LACP trunks. After a little while we seem
> to have a nice broadcast storm circulating in this one-way regulated
> ethernet highway :-) (So we right now keep both the interface "down"
> manually to break this loop).
>
> Relevant parts of the configuration from one of the routers/switches, A
> in this case - the others are configured similarly, in this one Trk1
> is A-C and Trk2 is A-B (the backup link)):
>
> trunk 1-2 Trk2 LACP
> trunk 3-6 Trk1 LACP
> spanning-tree
> spanning-tree Trk1 priority 0
> spanning-tree Trk2 priority 1
> spanning-tree config-name "IFM/Core"
> spanning-tree config-revision 1
> spanning-tree priority 1
>
> A related question - is is a good or a bad idea to enable
> "link-keepalive" (UDLD) on the links that make up these trunks?
UDLD is typically used to detect "unidirectional links". If you
think of the layered model, it runs below xSTP, so xSTP only
considers a link up if UDLD reports that both directions are
working fine. If your ports are directly connected and you are
running LACP, I see marginal benefit in using this feature.
It tends to be more useful if you have repeaters between the
switches because then you have multiple hops between switches
and a failure in one hop may not be seen by both switches, so
having UDLD would help detect a problem immediately.
Anoop
| 
Yahoo! 
Google 
Windows Live 
del.icio.us 
digg 
Netscape 
XML site map