My Amazon Elastic Compute Cloud (Amazon EC2) instance running Windows loses network connectivity one hour after boot up.
Time synchronization issues might cause the network adapter IP address Dynamic Host Configuration Protocol (DHCP) lease to expire. This issue only happens on Windows instances running in a virtual private cloud (VPC) when the time zone of the instance is changed to something other than UTC.
At boot up, the EC2 Windows instance gets the real time clock value from the underlying host BIOS. The clock value assumes it is in the time zone configured at the operating system (OS) level when it's really in UTC. After boot up, the Windows Network Time Protocol (NTP) client starts, syncs with the NTP server, and corrects the time. The time difference causes the DHCP client to stop renewing the lease. When the lease expires after one hour from boot up, the instance becomes unreachable because the instance cannot renew the IP address.
- Add the RealTimeIsUniversal registry key to the instance. The RealTimeIsUniversal key configures Windows to synchronize time settings with the system clock instead of UTC time.
- Update Citrix PV drivers to Amazon PV drivers. The Citrix Xen Guest Agent service can cause issues with time synchronization, so it is a best practice to use Amazon PV drivers instead.