Comware

Hi!  Here is my problem.  We have several wiring closets with 3COM 4400 switches in them in a stack configuration (2 or 3 switches at most).  We are seeing with 3COM Network Supervisor that many of the units are removing themselves from the stack causing sporadic connection issues.  We isolated one closet and replaced ALL of the hardware... new switches, new stacker cards and cables, new power cables... worked for about a week, then started removing themselves from the stack again.  What could be causing this?

Thanks in advance for any assistance.

Are you using the latest firmware version? Are the electrical infrastructure components working fine?

Regards

Firmware version is 6.13 and the power infrastructure is fine according to our Maintenance dept.

Also, the closets are plugged into UPS's.  If it was a power failure of some kind, would not both switches remove themselfs from the stack?

We are a school district with 5 buildings and we have been having this problem district wide.

It's really weird. Do you know if some special feature was enabled before these issues appear? Maybe some feature is forcing the switch to leave the stack, as an overhead or hig CPU utilization...try to check the system status of the switches that leaves the stack just after the issue occurs.

HTH

No special features have been enabled that may have caused this issue.  It has been happening for over 6 months now, we have just been putting up with the outages until we can figure out what is causing them.  We have changed out switches, stacking and extender kits, cascade cables, evrything and the problem still occurs with randomness.  It cannot be isolated to any one building or closet.  They are removing themselves all over our network. 

To give a little more background, all of the satelite locations run to a main 3COM 4950 switch via fibre.  We have a flat 10.0.0.0 network which I am sure is not helping with the traffic load.  Could this be causing switches to remove themselves though?

I can see a few possibilities that could force this behavior in the switches: some STP issue in the stacks or in the topology; some configuration that is missing in the stack configuration; a software bug in the running version. I don't know the stack configuration of the switches you have, but even in the latest hardware versions this problem occurs concerning the stack configuration and STP issues.

I would try the following steps:

1. Run a network discovery using 3Com Network Director and generate some topology and misconfiguration reports in order to check if there is something that is missed on the configuration and topology.

2. Create a lab using the same software version, using a topology similar to the environment you have, and check the stability of the stacks. If does not occur any problem, check the traffic flow. If the problem occurs, perform a downgrade one version back and perform the tests again.

3. If you have a flat network, consider to create new segments, optimizing the traffic and reducing the size of broadcast domains. Maybe these switches are learning a large amount of MAC addresses, what will causes a crash in the stack removing the switch from it.

HTH