Strange network issue: Certain clients getting no traffic.
Setup is as follows. All switches, servers and printers will respond to pings, and don't have any acls set that would block the ClientA or ClientB. Spanning tree is turned on and the setup has been stable and reliable for year+.
ClientA can ping Switch1, Printer1, Switch2, Printer2 and the BackboneSwitch, but cannot ping the Servers or other switches beyond BackboneSwitch. If netbooted, it gets handed an IP from DHCP Server, but can't load an image (WDS or Thinstation).
ClientA & ClientB -- Switch1 -- Switch2 -- BackboneSwitch -- Server(s)
| | |
Printer1 Printer2 OtherSwitches
ClientB can ping everything above, including the Servers and OtherSwitches plugged into the backbone, it can also load an image by netbooting.
If I take ClientA and plug it into BackboneSwitch directly, it can ping the Servers and OtherSwitches and netboot. If I then plug it back into Switch1, it can now ping the Servers and OtherSwitches and will netboot happily.
This is happening to a handful of clients (laptops and one desktop so far). With Intel and Broadcom nics of different makes, both 100BaseT and 1000BaseT. Switch1 and Switch2 have both been reloaded to a simple (IP + login details) known-good config just to eliminate them as a possibility and the issue occurs on other paths to the Backbone that avoid Switch1 and 2.
As far as I can tell. the issue is with the individual clients (unlikely, affects diverse Linux and Windows clients so not patch-related and clients are known-good) or the backbone switch. The issue occurs whether the intervening switch is managed or unmanaged. Switches are Cisco and Netgear.
I suspect something like MAC aging not working correctly, but might be completely wrong. Any ideas?