STP, RSTP, management issues

[Jim Lucas_1]

We are experiencing problems communicating (telnet or console) with the management of several HP Procurve 2524's. It appears that when RSTP is enabled, the problem is the worst (communication is so slow that you cannot navigate the menus - must use CLI). With STP enabled, they respond a little slower than average (but tolerable). And with spanning tree completely disabled, they respond quite well. Does RSTP not play well with other devices that run STP only, or does this sound like a firmware problem.

One thing I have found in common with two of the devices having difficulties is that they both have the 10/100/1000TX module and are directly connected to another 2524 with the same module set at 1000-Auto. They also both have a 100MB Fiber Module which uplinks to backbone equipment. Ironically, we are not experiencing any difficulties communicating with other 2524's uplinked to these.

Please forgive my ignorance, I'm trying to learn about these issues, but I would appreciate any advice.

[Jim Lucas_1]

Shortly after posting this question I ran across the discussion regarding LACP at the following link:

http://forums1.itrc.hp.com/service/forums/questionanswer.do?threadId=223025

I disabled LACP on all ports and now the response time is GREAT - even with RSTP enabled. However, I would still like to know why. Something does not seem right, since other procurves work fine with RSTP and passive LACP enabled.

[IT-Enheten_2]

Agree! Have had the same problem ourselves. Turned off LACP - Problem solved! Another issue is when having RSTP in a ring the convergence time appears to be faster with copper then with fiber!

[Jim Lucas_1]

Well, unfortunately the problem resurfaced with passive LACP off. I've only tested this with one of the switches having difficulties. I'm going to try it with some others and see if it helps. I really am not seeing anything in the logs to indicate that any problems are occurring. I spoke with a technician from HP, we are also looking at the possibility of wiring problems, however, I am not seeing any statistics to indicate that any ports are failing to receive or transmit packets.

[Ralph Bean]

Jim -

Could you post your 'show tech' output?

Ralph

[Jim Lucas_1]

I could, but would rather not include everything for the following reasons. 1) it is pretty long for a posting, and 2) there is some network information I don't want to publicize.

If you could pinpoint areas of interest, I can post these except I would edit out any IP addresses.

[Ralph Bean_2]

Hi Jim -

Since you can specify a file attachment, size should not be a problem.

But to answer your question my preferences, in order, are:

1. Remove the sensitive information and post the rest as an attachment.

2. Post the config, event log, and port counters portions of the 'show tech'.

Regards,
Ralph

[Jim Lucas_1]

Ok, here is the entire capture output of show tech, with all IP & Hardware addresses edited out.

Just an FYI, you will notice several configuration changes in the event log. These were things I was doing while trying to figure out the problem.

As I'm writing this, I was trying to disable ports and discovered that when I disable the 100/1000TX module (port 26), the switch starts responding properly. CPU Util drops to around 18% You'll notice some issues in the log regarding errors with the 100/1000 module. I am going to try to reseat that and see if it clears anything up. Interestingly enough, as I mentioned in the origingal posting, I've experienced the same problems with other Procurve 2524's cross-linked in the same manner.

And, as I'm thinking of it... I may have assumed that the 100/1000 module supports MDI/MDIX. I have been using straight through cables instead of cross-over cables. I'll need to check the manual to see if 1000TX works in this fashion.

I'll follow this line of thinking and see where I wind up...

[Jim Lucas_1]

I've since re-read the Manual shipped with the 100/1000TX module. It claims that MDI/MDIX is supported. I was using a straight through cable - as I've been with all other situations with 2 procurves connected together with these modules. I tried stepping down both switches to 100-auto and subsequently 100-FullDuplex and I cannot get any link between the switches. I'm beginning to wonder if one of these modules is faulty. Next step, I'm checking this theory with a spare new module I have...

[Ralph Bean_2]

Hi Jim -

I suspect you are on the right track.

By the way, did you notice the event log message, "W 12/19/03 19:10:51 stp: BPDU has wrong Protocol ID and version from port 26"? What type of device is attached to the 2524's port 26 and how is (R)STP configured on it?

To Jim and any kibitzers, Happy New Years!

Regards,
Ralph