Request Timed out

[Michael_373]

Hi all,
I have 7 5308xl switches in 2 building. We are using meshing for the 3 main switches and truncking on the rest. I am trying to ping a server in another building and at a certain point, I get a request time out at every 10 or 15 pings. It can go 50 pings without a problem, and then oups, request timed out.

Anyone know where I should look? I have checked everything and I have also updated my firmware to 8.01.

Thanks all,

Michael

[Stuart Teo · Accepted solution]

Hi Michael,

A few questions I have. You mentioned that you are using Mesh to interconnect the 7 x 5308xl. How many of these 7 are actually in the mesh domain? What speed are these links? 100 or 1000mbps? Would it be possible for you to draw us a diagram?

[Michael_373]

here is the plan... the problem is when I go from 01swt101/102 to 01swt106/107/108.

Hope you understand the diagram... it's very detailed..

Yhanks again

[Stuart Teo]

Do you have STP/RSTP/MSTP turned on?

I've encountered both bad fiber jumpers and/or bad ports on the FCB that have led to similar symptoms. Have you checked end to end optical signal strength?

What GBICs are you using? SX, LX or LH? Are they nailed down to 1000fdx or Auto?

Where is the pinger and where is the pingee?

Trunk mismatch. Do you have the same trunking protocol set up on both ends? Static? PAgP/FEC? LACP?

[Michael_373]

We are using STP.

I haven't checked the signal yet...

The GBIC's are LX and they are all pinned down to 1000Fdx.

We are pinging from 01swt101 to 01swt106/107/108

For the trunks, we are using Static for all of them.

Hope this helps...

Michael

[Stuart Teo]

I would really check the fiber path. Also try this. Increase the ping packet size starting at 32 (default) and eventually moving up to 3200. See how that affects the drop rate. It'll give you a good clue as to whether the fiber path is a candidate for suspicion.

[OLARU Dan]

I had a similar problem with a PC hooked to a switch through a 110-meter horizontal UTP cable. It all went away after replacing the horizontal cable with a 90-meter one.

The symptom pops up when you try to ping the server from any location (server cable)? Or only from one location (cable to that location)?

Ethernet specs (including FastEthernet) require maximum 100 meters between PC's NIC and switch port.

[Michael_373]

I just had the fiber checks and all is 100 % ok.
I had another question, In the STP configuration, the switch priority is set at 32768 for all switches. Are they supposed to be at a different priority?

Thanks again.

[The Real MD]

Unfortunately I cant open the visio diagram, but is it possible you can run multiple traceroutes to see if there is any deviation in its route.

Regards

~Martin.



PS. dont forget to allocate points

[Michael_373]

how do I run multiple traceroutes? I tried and it's pretty direct from point A to point B.

[OLARU Dan]

Michael,
about STP configuration:

32768 is the default value of priority - same value on all switches. But if you have a switch that is at the center of your star topology and you want it to be the root bridge of the STP tree, you can set this switch's priority to 16384 and it will become the root bridge during STP elections.

Hth,
Dan

[Michael_373]

If I change the priority on one switch, does it have to be reboot? Do all of them have to be?

[Michael_373]

Can Flow control be activated on meshed ports, If so, do I have to reboot? Do all switches have to have the same flow control?

Thanks again

[Michael_373]

Sorry about writing so much...

I just tested from the CLI of one switch and pinged another switch directly. I get a couple of time out requests. Is this normal from one switch to another?

Thanks again

[Stuart Teo]

Hi Michael,

Since you run the 5300xl with 8.01, I'd strongly suggest that you run RSTP rather than STP. STP has certain deficiencies due to its specification (as opposed to implementation). The RSTP specification fixed some of these deficiencies. RSTP also moves the ports to forwarding mode faster.

Changing the switch priority does not require any switch to reboot. The BPDUs containing the new priority informating will automatically change the network topology.

On choosing the root, the router is normally the best root. Of course I'm talking about the newer switch-routers like the cisco 6509. If your router is not STP capable then you should choose the switch that is connected to the router to be the root switch.

One good thing about RSTP on the 5308 is that it automatically falls back to STP on a per port basis. E.g. cisco routers don't support RSTP so the port connecting the 5308 to the cisco will step down to STP automatically after a few BPDUs.

But these are not hard and fast rules. To fully understand how to design your spanning tree, you should read http://www.cisco.com/en/US/tech/tk389/tk621/tk5/tech_protocol_home.html

Read up also on port cost. Of all the spanning trees I've built, I've never tweaked port costs before (but it never hurts to understand how to tweak it, just one more weapon in the arsenal). All I've ever done is to modify the switch priority to force a primary root and a backup root.

[Stuart Teo]

Assigning a port to the mesh or removing a port from the mesh requires rebooting.

If a port is set to AUTO then flow control will be automatically negotiated whether or not that port is a mesh port.

Missing the 1st or 1st two ping(s) when you ping from one switch to another is common. At least for me. I see this on my swithces pretty frequently.