We're having problems which might be similar to what you're seeing, although in our case it's not on our ERS 4500's but on our ERS 2500's. It's currently up to 6 ERS 2500's which were spewing out the following error message on their serial console (nothing in the logs or when we connected through telnet/SSH, if we could still connect that is):
...
Tucana MMU DRAM CG0.M1 CRC error at 0x000010ab
Tucana MMU DRAM CG0.M1 CRC error at 0x00001102
Tucana MMU DRAM CG0.M0 CRC error at 0x00000ff5
...We RMA'ed all 6 of them and are in the process of getting fresh ones from Avaya. The frustrating part is that we cannot check this remotely and that we essentially have to wait until the connectivity to the switch dies (and our monitoringsystem notifies us) or until the users behind the switch start complaining because of (very) bad network performance. My guess is that Avaya have gotten a bad batch of DRAM chips from a 3rd party vendor and some of those switches are (slowly) dying and exhibiting this behaviour (we currently have 70+ ERS 2500's on our network so I'm really keeping my fingers crossed that we're not facing a massive swap operation).
A couple of months ago, when I saw this the first time, I also couldn't make any sense of it but we immediately needed those ports so I just tried a reboot which seemed to clear up these error messages (although it failed the internal loopback test on rebooting) and left it at that because we already had plans to clean out that wiring closet. A couple of weeks ago we actually did the makeover of the wiring closet in question and swapped out the bad switches (2 of them which still failed the internal loopback test) and started an RMA for them. So in hindsight these switches actually still worked (kinda I guess) and we never received any more complaints but ofcourse we kept a close eye on it.
The two things we currently do to test this is to first try a reboot and see if any of the POST checks fail, if this is all OK but we still suspect the switch I just create a loop on it (in a closed lab environment) to easily generate some traffic after which some of them again started to spew out the above error messages on their serial console (so it might also be traffic and/or buffer/queue related).
Where I'm getting at with all of this is that if I were you I'd keep a close eye on that switch and if there would be any doubt just replace and RMA it. This networking stuff is already hard enough sometimes, we really don't need flaky hardware making it even harder
.
