Switches, Hubs, and Modems
cancel
Showing results for 
Search instead for 
Did you mean: 

rstp takes about 30s to be ready... ?

 
Highlighted
Frequent Advisor

rstp takes about 30s to be ready... ?

Hi!

I've two Brocade XMR and about 30 HP Switches which run rstp. If the root bridge fails (one of the XMR's) the whole network takes up to 30s to be ready again.

Is this normal? I thought at least with rstp it should be really fast.

When this happens i see a lot of messages like this:
"FFI: port 7-High collision or drop rate. See help."

The config of an HP Switch looks like this:
spanning-tree
spanning-tree 1 bpdu-filter bpdu-protection
...
spanning-tree 48 bpdu-filter bpdu-protection
spanning-tree priority 15 force-version RSTP-operation
loop-protect 1-48

Uplinks are at Port 49 and 50. 1-48 are clients.

Stefan
48 REPLIES 48
Highlighted
Frequent Advisor

Re: rstp takes about 30s to be ready... ?

Do i have to set something like admin-edge or have i to remove loop-protect or spanning tree at all on the ports to the clients?
Highlighted
Esteemed Contributor

Re: rstp takes about 30s to be ready... ?

Hi Stefan

Did you configure the one XMR with say priority 0 ad the other with priority 1.

Secondly what kind of hp switches is it and what software version is on them.

KR
Gerhard
Highlighted
Frequent Advisor

Re: rstp takes about 30s to be ready... ?

> Did you configure the one XMR with say
> priority 0 ad the other with priority 1.
One has Prio 0 and the other one Prio 8192.

> Secondly what kind of hp switches is it and
> what software version is on them.
2848, 2824 and 2650-48

2848 => 10.73
2824 => 10.73
2650 => 11.54

Does it make sense to use admin-edge? And if which ports should be configured as admin-edge? Could it make sense to disable spanning tree at all on the edge switches? Does it make sense to have loop-protection AND spanning tree on?
Highlighted
Honored Contributor

Re: rstp takes about 30s to be ready... ?

You haven't got the thing with the routers runnning 802.3w draft 3 like last time around?
Highlighted
Frequent Advisor

Re: rstp takes about 30s to be ready... ?

No - we buyed new 2010 XMR routers :-) which only know 802-1w protocol.
Highlighted
Esteemed Contributor

Re: rstp takes about 30s to be ready... ?

Hi Stefan

First item to note. A spanning tree root-failure is about the worst thing that can happen to a spanning tree. This can potentially lead to a significant convergence time.

Edge-port does not really help with root bridge failure and new bridge election.

Let me digress for a bit and then come back.

I see you have bpdu-filter and bpdu-protection on your edge ports. Those two features are sort off mutually exclusive. I would suggest you only use BPDU-protection. ( Except if you have a client on the port that does not like spanning-tree BPDU's, I have seen some very old systems crashing if they receive an STP BPDU.)

If a port has BPDU filtering configured it does not mean it will skip the blocking state, it means it will ignore BPDU's after the initial blocking state and always stay forwarding.

It is a good policy to configure edge-ports as admin-edge though.
-- To see if this might help you in your situation, would you mind posting the output of show span and show sys for a few switches(one of each model), I just want to see the state of the ports.

The above note is where admin-edge port might help. Since admin-edge implies it will skip the initial blocking state. Lets look at the output first though.

Having said the above.
2824/68 => 10.73
2650 => 11.54 ( Is this not a 2610 ? )
I just want to verify the versions and models above to be sure. :).

-- FYI NOTE --
BPDU-protection is for checking if you receive BPDU's where you are not expecting them, and then protecting that port by shutting it down.

BPDU-filtering is for throwing away BPDU's and that is if you have a sport where you do not want BPDU information to be shared circa a point to point routed link between separate switch domains.

-- END FYI --

In the end yes RSTP is very quick but you have to get the configuration good and right, as there is many options.

HTH
Gerhard
Highlighted
Frequent Advisor

Re: rstp takes about 30s to be ready... ?

First i would like to thank you for your detailed answer.

> This can potentially lead to a significant
> convergence time.
OK good to know - but how long is this normally?

OK first things i should think about are:
1.) only use bpdu-protection
2.) configure each port where only a client/server is attached as admin-edge
3.) 2650 => 11.54 ( Is this not a 2610 ? )
i'm sorry i meant 2610 my fault.

4.) i've attached a .txt file with sh sys and sh span from different switches and some more explanation which switch is connected to which one.

> In the end yes RSTP is very quick but you
> have to get the configuration good and
> right, as there is many options.
I'm sure my config isn't the best.

Stefan
Highlighted
Esteemed Contributor

Re: rstp takes about 30s to be ready... ?

Hi Stefan

I would make the following initial recommendations.
1. For all interswitchs links(trunks and ports) to disable the auto-edge functionality on them with:
no spanning-tree PORTNUMBERS auto-edge-port

2. For all the edge ports. I would personally remove BPDU-filtering
no spanning-tree PORTNUMBERS bpdu-filter

3. For all the edge ports enable admin-edge
spanning-tree PORTNUMBERS admin-edge-port

Be aware that some of you switches port 48 seemingly as interswitch links. It is a bit bad if you interswitch links are admin edge as this might lead to transcient loops.

Looking at your data, I have noticed some buffer misses, this might be attributed to transcient loops.

Watch out for loop-protect on inter-switch links.


Also lastly make sure you Brocade ports linking to the switches is not auto-edge or admin-edge ports. ( I do not know the command for that one or how to verify it in the config).

Then what I would suggest once you have everything configured through the whole LAN, give it a test. If it improves good, if not post the output of show logging for a few switches with redundant connections to the core.

The recovery itself giving a number of expected seconds is not really realistic it depends on some factors most of all the diameter of your network.

HTH
Gerhard

Highlighted
Frequent Advisor

Re: rstp takes about 30s to be ready... ?

Hi

i'm trying to put your suggestions into my config. But i've some problems:

> 1. For all interswitchs links(trunks and ports) to disable the auto-edge functionality on them with:
> no spanning-tree PORTNUMBERS auto-edge-port
You mean "admin-edge-port"?

There is no auto-edge-port config.. ?

2.) Shouldn't i also disable LACP where it is not in use? I've seen in log that the switches are also checking for LACP on each up / down.

3.) Does loop-protect work fine with Spanning-Tree?

I've this on one switch:
Loop Loop Loop Time Rx Port
Port Protection Detected Count Since Last Loop Action Status
---- ----------- --------- ---------- ---------------- ------------ --------
1 Yes Yes 1 6d,7h,2m,11s send-disable Up
2 Yes No 0 send-disable Up
3 Yes No 0 send-disable Up
4 Yes No 0 send-disable Up
5 Yes Yes 1 12m,35s send-disable Down
6 Yes No 0 send-disable Down
7 Yes No 0 send-disable Down
8 Yes No 0 send-disable Down
9 Yes Yes 1 12m,35s send-disable Down
10 Yes No 0 send-disable Down
11 Yes No 0 send-disable Down
12 Yes No 0 send-disable Down
13 Yes No 0 send-disable Down
14 Yes No 0 send-disable Down
15 Yes No 0 send-disable Down
16 Yes No 0 send-disable Down
17 Yes No 0 send-disable Down
18 Yes No 0 send-disable Down
19 Yes No 0 send-disable Up
20 Yes Yes 1 12m,30s send-disable Down
21 Yes No 0 send-disable Down
22 Yes No 0 send-disable Down

But i don't understand how it could detect a loop on f.e. port 5 where the port is down?! Also i don't understand why it detects a loop on port 1 and the port status is still up?!

Thanks again.

Stefan