Strange switch behaviour in VLAN network

Hi all,

I'm experiencing some weird problems in our switched network and am really stuck there. Our configuration is the following:

We've got an Extreme Summit1i 8 x SX GBit basic layer 3 switch (OS Version 7.2) working as "backbone switch", to which all our other switches are connected. Our floor switches are stacked 3Com

3300s release 1 with current firmware, the servers are connected via a Summit48i (also basic layer 3).

We have configured VLANs on all the 3Com switches and the Summits and disabled STP (all VLANs are located on one floor switch only). The two Summits exchange their VLAN information via the proprietary EDP (extreme discovery protocol) and their VLAN tables look correct. This configuration was working fine for two years until recently.

Suddenly all the switches started to drop packes for alternating destinations without apparent reason. For debug reasons I connected one of the servers directly to the backbone switch and saw that from time to time the server wasn't able to ping the ip address of the switch (i.e. its default gateway) and couldn't reach into some floor networks, but at the same time a continuous ping from a computer on the Summit48 to this server ran without any drops.

To illustrate that, here are some excerpts from the configurations:

------- Summit1 default VLAN --------- VLAN Interface[0-200] with name "Default" created by user Tagging: 802.1Q Tag 1 Priority: 802.1P Priority 7 IP: 192.168.0.1/255.255.240.0 STPD: s0(Disabled,Auto-bind) Protocol: Match all unfiltered protocols. Loopback: Disable RateShape: Disable QosProfile:QP1 QosIngress:None Ports: 8. (Number of active ports=7) Flags: (*) Active, (!) Disabled (B) BcastDisabled, (R) RateLimited, (L) Loopback (g) Load Share Group Untag: *1 *2 *3 *4 *5 *6 7 *8

---------------------------------------

-------- Summit1 VLAN ITS ------------- VLAN Interface[10-20a] with name "ITS" created by user Tagging: 802.1Q Tag 51 Priority: 802.1P Priority 7 IP: 192.168.51.254/255.255.255.0 STPD: s0(Disabled,Auto-bind) Protocol: Match all unfiltered protocols. Loopback: Disable RateShape: Disable QosProfile:QP1 QosIngress:None Ports: 1. (Number of active ports=1) Flags: (*) Active, (!) Disabled (B) BcastDisabled, (R) RateLimited, (L) Loopback (g) Load Share Group Tagged: *2

----------------------------------------

The switch with the clients in ITS VLAN is indeed connected to port 2 of the Summit1.

-- 3Com floor-switch 192.168.0.243 default VLAN --------- VLAN ID: 1 Local ID: 1 Name: Default VLAN

Unit Ports

1 none 2 24, 25

Unicast Frames: 902 Octets: 4296726 Multicast Frames: 2796 Broadcast Frames: 55253

----------------------------------------------------------

----------- 3Com floor-switch 192.168.0.243 ITS VLAN ------------ VLAN ID: 51 Local ID: 6 Name: UHD4

Unit Ports

1 9, 10, 11, 12, 13, 14, 15, 16 2 25

Unicast Frames: 90195 Octets: 48910162 Multicast Frames: 0 Broadcast Frames: 253

-----------------------------------------------------------------

STPState on the 3Com is "disabled".

What really startles me is that both the Summits and the 3Coms show the same behaviour, you can connect to them, ping them, and without visible reason they become unavailable for some time, but they still forward packets for other destinations.

I'd be really grateful for any hints where to look, and I thank everyone in advance who reads through my rather lengthy description.

-Chris