Horde agents has a ~40s “deadtime” at the beginning of the first node executed on an agent. This is evident from Horde UI and when watching things happen in real-time or via OTel, but not in the logs; see attached screenshot.
After some investigation, I’ve determined that this is due to JobDriver attempting to set AWS-related environment variables:
`// JobExecutor.cs
// TODO: These are AWS specific, this should be extended to handle more clouds or for licensees to be able to set these
newEnvVars[“UE_HORDE_AVAILABILITY_ZONE”] = Amazon.Util.EC2InstanceMetadata.AvailabilityZone ?? “”;
newEnvVars[“UE_HORDE_REGION”] = Amazon.Util.EC2InstanceMetadata.Region?.DisplayName ?? “”`These properties call AWS SDK methods which makes HTTP requests that only work from within EC2. With built-in retries and caching, when not running within AWS, attempting to fetch these two values causes ~40s of inactivity on the first node running on an agent, and ~16s for every subsequent node on the same agent.