HP 4018GL Switch Port Problem

[Paul Lambert_2]

Hi all,

I'm having a problem with some of my 4108gl switches. I'll get a complaint that a device is not getting any network connectivity. When I look at the switch, there will be a link light, but the activity light flashes really slow. I'll plug the patch cable into another port and the device works fine. On several occasions, when a port goes bad like this, several other ports on the same module will go bad and usually all of the bad ports are physically "close" to each other (e.g. ports 17,18,23,24).

I'll eventually replace the module with one of our spares. When place the module with the "bad" ports in my test switch, all of the ports work.

This has happened on switches in several building on different subnets. Any ideas?

[OLARU Dan]

If you get the same symptoms with the known-to-be-good module, then it is most likely the switch backplane that needs changing. I have this problem, too, in a 4108gl where some ports in some modules work extremely slow, and they're buildup impressive amount of errors - and I already changed the modules with good ones.

4108gl was a bad design from the very beginning (at least compared to 4000M/8000M series, which are extremly robust, with high MTBFs)- as part of lifetime warranty HP should provide their customers with re-engineerd backplanes and modules when these fail: they could easily win more clients in the ProCurve camp than by using the current practice/policy of providing replacemnts that were just "refurbished"!!!

[Matt Hobbs]

I don't know about that. I've never had any problems with my 4100's, but I've had a few fan failures in the 4000M's (the new fan is a better design though). I don't think HP would offer a product with a worse MTBF when offering a lifetime warranty. You could actually get these figures off HP (although getting the 4000M figures may be difficult now).

In regards to this problem description though, it's not something I've come across before. Which type of module is it occurring on?

[Paul Lambert_2]

All of the modules with bad ports were J4862A 10/100-TX module. We have a mix of those and J4908A 20 Port 10/100/1000-TX on our campus.

HP has been very good about replacing the modules, but I'm suspecting they are not the problem.

[Ari Yrjölä]

I've seen this same problem several times on our 4108gl switches with 10/100 ports. When it happens, there's always two syslog entries with this port going off-line without going on-line in between. Port link light shows that the link is up, but switch software does not register it any more. This zombie port will come back alive after switch reboot or physical module reset.

[Paul Lambert_2]

I had another port go "bad" the other day on one of our switches. I looked at the error log from that switch and saw the same error you described:

I 04/16/07 23:38:29 ports: port E24 is now off-line
I 04/16/07 23:38:29 ports: port E24 is Blocked by LACP
I 04/16/07 23:38:29 ports: port E24 is now off-line

Rebooting the switch or resetting the module is obviously not a great solution.

I'm not familliar with LACP. Could that have something to do with it? The LACP setting are just the defaults.

[William Ey]

Hi There,

I have a 4104GL and am experiencing exactly the same problem, down to the log entries and the fact that the ports are in close vacinity.
This switch has been operational for going on 3 years now and this is the first time that i have experienced this problem.

Did you every find a resoultion?

Thanks in advance for any help that you can offer.

Regards

Will

[joe_butler]

I have found that LACP can slow down ports as they first come up. if this port is flaping then LACP will add a delay each time.

Personally I turn LACP off on all ports, the only time i have ever had to put this back is when a dual nic server is plugged into the switch and then LACP can be used to create a dynamic trunk.