[CentOS] ECC memory errors

I started to receive this kind of messages a few days ago on one of my
servers:

Message from syslogd@ at Mon Apr 29 08:02:55 2013 ...
server1 kernel: EDAC MC0: UE row 0, channel-a= 0 channel-b= 1 labels "-":
(Branch=0 DRAM-Bank=0 RDWR=Read RAS=0 CAS=0, UE Err=0x2 (Aliased
Uncorrectable Non-Mirrored Demand Data ECC))

I've never had ECC memory to fail on me before, so now I am wondering the
following:

* The server is running CentOS 5.7 and is acting as Xen dom0. Is there any
possibility this could be a kernel issue and upgrading would help, or would
upgrading at this point just cause more trouble?

* Is there now a possibility that my data can get corrupt: should I
shutdown the server as soon as possible or can I keep running until I
replace the memories?

* This server has been running for several years in a datacenter without
problems: what are your experiences, are these kind of problems most likely
caused by a failing motherboard or the memories?

Regards,
Peter