error causing unstability in a mesh network

neverland56 · ‎02-27-2007

Dear all,

I currently have 4 cisco 6006 switches (A,B,C,D) in a mesh network (except C & D not link together) and 3 Cisco 4510R switches (E,F,G) all links to C (Po1) & D (Po2). switch E,F,G do not interlink each other.

While I was doing a UAT test, I do a continuous ping from the client's side of G to the outside world. Everything goes fine when all equipments is up. But when I power down switch C, I found that sometimes I can ping but somwtimes it will show "request time out." I was wondering what went wrong to the networks because from what I understand, in such a mesh network, by right unstability should not be occurring. I do not know what is wrong and how should I go about with the triubleshotting.

Hope that you guys will be able to help me. Thanks

Jon Marshall · ‎02-27-2007

Hi

You don't say whether or not you are connecting all your switches at layer 2 but assuming you are.

Based on what you have described this looks like a spanning-tree issue. You do have layer 2 loops in your network so one link will be active for the client vlan and one will be blocked.

If the active link is from G to D then when you power down switch C it should have little effect as G is already forwarding to D.

But if the active link is G to C then when you power down C spanning-tree has to find another link which will be G to D. This can take some time.

Which spanning-tree are your running ?

PVST+

RPVST+

MST

Also when the ping times out does it come back eventually and if so roughly how long ?

HTH

Jon

neverland56 · ‎02-27-2007

Hi,

all my switches are connected as Layer 2. My spanning is running PVST+. The ping times out will come back eventually but after a little while (around 3-5min) it will times out again.

Timed out time is ard 10s-20s.

Thanks

Jon Marshall · ‎02-27-2007

Hi

Okay. PVST+ can introduce a delay of up to 50 seconds on the failure of a link so 20 seconds isn't too bad :-)

Rapid PVST+ would significantly reduce this as long as all your switches are capable of running it.

The 3 - 5 mins timeout could well be switch C coming back up. When it comes back on line it now introduces potential loops into the networks so spanning-tree has to recalculate again.

HTH

Jon

neverland56 · ‎02-27-2007

Hi,

ok. Thanks a lot. But I had another inquiry. I had powered down switch C and observed for around 15min. The sequence of the ping is as follows:

ping (ard 5min) --> timed out (ard 20s) --> ping (ard 5min) --> timed our (ard 20s) --> and it continues the same pattern for the whole of 15 min (before i power on my switch C). So I was just wondering, is the above scenario normal?

Thanks

Jon Marshall · ‎02-27-2007

Hi

No it's not normal. Is there any chance that the link from G to D is getting overloaded now it has to take all the traffic that was previously sent over both links before ?

How are these links setup. Do you load balance your vlans over these links by setting spanning-tree root and secondary on a pair of your 6006 switches ?

Jon

neverland56 · ‎02-28-2007

Hi,

I'm sorry that I do not know that whether load balancing was properly set on the the 6006 switches as the previous network engineer who design this network had left and did not leave anything behind.

I had attached the config & spanning-tree file below and I do hope that you will be able to spare some of your time and help me take a look at the config so as to alighten me whether I should add-in/out some additional commands to help me solve this issue of unstability.

From what I understand :

- Root bridge is at switch A

- Ping test is issued at vlan 9 of switch G

Thanks a lot

Jon Marshall · ‎02-28-2007

Hi

Could you send the config for switch B as well. From a preliminary look switch A has been setup as the root for all vlans.

Where are you pinging to from the client on switch G.

Also can you send the output of the following from your switches assumming switch C is up.

Switch A, B, C & D - sh spantree 9

Switch E F G - sh spanning-tree vlan 9

Have to shoot off to a 2 hour meeting (joy !!!) but will try to get back later.

Jon

neverland56 · ‎03-11-2007

Hi,

very sorry for the long late reply. Pls see as per attached for the Switch B configs and also the output configs of the spanning-tree.

From the client on switch G, I was pinging to the following places:

(1) 2 servers on vlan 1

(2) 1 server on vlan 2

(3) 1 server on vlan 9

(4) the WAN network of the HQ of the company

Hope that you will be able to help.

Thanks a lot.