Catalyst 2950, 3500, and 2948 switches

jesusdamian · ‎06-04-2007

all with fiber connections. Each fiber connection runs to a single 3508G-XL.

We are having what I beleive is SPT convergence 4-5 times a day.

Portfast is enabled on the fiber ports, and it was my understanding that PortFast should be disabled on any port that connects to a switch or hub. Since the fiber connects to switches, should I have PortFast disabled on all fiber ports?

would that make a difference?

sundar.palaniappan · ‎06-04-2007

Hi

Enable portfast on all ports connected to end devices, like PC's/Servers etc. Disable portfast on all ports between switches. You shouldn't be seeing too much STP reconvergence if the network is quite stable. If the connection between switches are flapping that can cause STP to reconverge.

If you are still having problems can you elaborate what type of problem you are having and post the relevant output from the switch.

HTH

Sundar

mark.j.hodge · ‎06-05-2007

From your description I doubt it is spanning tree. If I read you correctly you have a Star like topology with the Hub swith (3508G-XL) connected via a single Gb fibre to each of the Spoke switches (2950, 3500, 2948XL).

Spanning tree convergance will only take place if you have multiple links, either there are two connections between a Spoke and the Hub, or two Spokes are connected together.

As previously stated, the switch-2-switch connections should not have PortFast enabled, on either device.

Without any further information, I would suspect a broadcast/multicast issue, are there any particular activities being performed on your network at the time you have issues. Imaging tools (Ghost etc) are regular sources of such issues.

** Please rate posts if helpfull **

jesusdamian · ‎06-05-2007

We arent doing anything more than your standard network operations. 20 people connected to a client/server app and the rest are authenticated, logged in, and doing various functions.

What happens is, 3-8 times a day, all network activity stops across all switches for about 45 seconds, then, with no user intervention, everything comes back online.

But by then, all users have to re-log into the network.

and yes, its a star topology with a central 3508G-XL connecting it all.

mark.j.hodge · ‎06-05-2007

If there are no loops in the network, then it realy can't be spanning tree.

Check the logs on the Hub 3508G-XL immediately after an event, this should give you a better indication of the cause.