network issues - inconsistent pings across network

[Nick Wickens]

Hi All

Had some problems with our network switches this morning which is partially resolved but the issue I am left with is confusing.

From our K370 I can quite happily ping any address within our 171.10 subnet. However from our K450 I can only ping certain addresses either way. Ie from the K370 I can ping 171.10.70.230 and vice versa however the K450 can't ping this address and neither can the machine at this address ping the K450. eg

171.10.20.1 can ping 171.10.20.6 and vice versa.

171.10.60.100 can ping 171.10.20.1 & 171.10.20.6

171.10.60.230 can ping 171.10.20.6 BUT not 171.10.20.1.

I have checked DNS out and the routing tables on both HP9000's are the same. Its not just the HP9000 kit though - some printers on 171.10.70 are fine but others are not.

If I pass my network connection over to my collegue he is unable to ping kit that I can when I use the same connection but we are both in the 171.10.60 range.

Any ideas ?

[Stefan Farrelly · Accepted solution]

Hi Nick,

It sounds like a network issue, either your
default gateway or router or switch are playing up or have their routing tables in a muddle. You probably need to reboot whichever one it is.

Thats all I can think of.

Cheers,

Stefan

[Nick Wickens]

Hi Stefan (and sorry I keep spelling your name wrong).

Had already decided that a reboot was the next valid action (its only a development system) and its no back up and ok.

Cheers

Nick

[harry d brown jr]

If you can't afford a reboot or the switches/routers, which wouldn't make sense since you are having some serious issues anyways, but most have the ability to clear arp cache. You also need to have them DOUBLE check their routing rules!

I know my network guys lose their minds when someone points their finger at the network, but ..it happens and they need to deal with it!

live free or die
harry

[Nick Wickens]

I had tried clearing out the arp cache with arp -d for those machines having the problem but without any luck and routing tables had not changed (in fact more or less identical to the other HP9000 server that did not have the problem.

The last fall back solution of the desperate man worked though - power it off/on. It usually does :@).

[Hartmut Petroll]

Hi Nick,
did you check the subnet masks on your machines?
What does the traceroute say?

Kind regards
Hartmut

[John Dvorchak]

It would be helpfull if you tell us the netmask on all systems and some traceroute outputs from the machines that work and don't work. They are prabably ok but if the Netjerk guys don't believe the problem is in the LAN then at least you have some amunition to plead your case that they messed up.

[Onan Coca_1]

Nick,

It looks like you lost some route statements in your router. I seen this happen with RIP.
Reboot your router or Flush your routes.

Onan

[Ron Kinner]

As a network guy I know the rule is The Network IS Guilty Until Proven Innocent. In this case I think it applies. I would recheck the masks on all boxes involved but your problem is not really consistent with the usual bad mask problem.

Try a linkloop test between your problem machines. You need the mac address of the other PC and you need to know which NIC it's going out on.

If you only have one NIC then I think you can get by with
linkloop -n 5 -v 0xMAC

where MAC is the address of the remote PC. Try it first using your own MAC address which you can get from lanscan. IF that doesn't work you need to use the -i option to identify the correct LAN to use. Then try it with the MAC of a machine you can ping so you know what you are doing. Now try one you can't ping. IF this works it's your problem and the Network is innocent. If it doesn't the Network is guilty.

If the Network is guilty then look for static MAC assignments which are no longer valid. Look for trunks down between switches. Check VLAN assignments. Verify that the STP root bridge is correct and if using VTP that they didn't replace a switch and use one which thought it was boss when it wasn't.

Ron

[Nick Wickens]

Thank you for the additional suggestions but the issue is now resolved. Turned out to be a duff switch module that did'nt reveal itself until we forced the spanning tree reconfig (pulled out a fibre cable) whereupon the fault light started flashing and the module "died".

My suspicion is that all of the machines that could'nt access our HP9000 were attached on that particular switch.

Interesting that there appears to be a mini-conflict between network people and other IT staff - working on small sites we all just muck in on network tasks with just the occasional light hearted jibe at how flakey microsoft security is :@)