RSTP Issue?

[TomHill]

Hi Guys,

I have a really strange intermittent issue that i'm having difficulty pinning down.

The issues are contained within one area of the network that consists of about 10 2650's and 1 5308xl in a ring topology. I was noticing pings being dropped internally, about 5%.

My first thought was high utilisation but I couldn't identify any saturated links as such so I assume other traffic is also similarly effected and we have had issues with IP phones not registering etc. that seems to be linked. The loop is 1gig with access ports 100Mb.

The first time we had the issue a couple of weeks ago we solved it through trial and error by tracing what we beleived to be the cause down to a SOHO wireless router that a user had recently plugged in. Unplugging the device instantly resolved the issue, i.e the packet loss dissipeared. However the issue has now returned today.

It's strange as this time I rebooted one of the 2650's and the problem went away. It just so happens that 2650 is both the stack master and spanning tree root (if that's the correct terminology, switch priority = 0). It's now been fine for a number of hours..

Does anyone have any ideas as to what could be causing this?

I was wondering if I should shift the stp root over to the 5308? Just a thought.

Thanks for reading.

Tom

[Pourl]

Hi I would make the most powerful and -central switch to the root. But I would check the uplinks for broken cable. Check the logs for crc and jaber errors check the interfaces.

A nice aid is a syslog server which collects all messages. So u can see where frequently errors occor.

CLI command:

logging <ip-address of syslog-server>

[TomHill]

Hi Pourl,

That sounds like a plan. Any Windows based Syslog servers you would reccomend?

Tom

[Pourl]

I use Kiwi Syslog, but I dont know alternatives. There are free Syslogs... try and error :)

[TomHill]

I have installed a trial of Kiwi and set up the logging on all of the switches in question. I also turned up Fault Detection>Log Network Problems to High in the web GUI.

I'm getting quite a few port x PD Invalid Signature indication, not sure if that's anything to worry about.

Tom

Edit

Issue Description:

The Event Log of a PoE-enabled HP ProCurve Switch may list the message:

ports: port B1 PD invalid signature indication.

Solution:

The message means that a device has been connected to port B1 which does not conform with the IEEE 802.3af PoE standard. The warning is informational, and requires no action by the administrator.

[TomHill]

I think it might be a bandwidth issue. Although the ring is 1gig we then connect this in to our core with 2x 100Mbps links in an etherchannel.

I can see via the web gui that one of these links is hitting 99%, the other about 25%.

If this is the case would I see the pings that are not getting through be recorded as dropped packets? The pings originate from the core.

sh interface x (interfaces of both of the etherchannel ports) on the HP side shows some drops but is not incrementing. Neither is the other side of the etherchannel which is a Cisco 3560...

Would ICMP packets be recorded?

I think I have the bits in stock to swap this out for a gig fibre circuit tonight but would like to prove this as the issue first if possible.

Tom

[Pourl]

The Interface records every paket no matter which protocol.

But u dont told about a cisco 3560 before?! Do u got a topologie plan which shows device-typ, links, stp-protocol, stp-prio?

[TomHill]

Ahh sorry, I didn't think it was relevent. In hindsight it couldn't have been more relevent. Upgrading the link to 1Gb solved the issue. It seemed the 2x100Mb links were the pinch point.

I'm not sure what counter the dropped ICMP packets were attributing to though, although I could see some dropped packets on the etherchannel it wasn't enough.

Tom