Yea, it is DNS caching the IPv4 and not listening to the TTL. In this case the DNS should be renewed after 600 seconds (10min) for 'api-us.roborock.com' CNAME and only 2 seconds for ELB A record.
Here is the dig command on that CNAME during an error with the expected IPv4s:
Here is the wireshark of Hubitat resolving INCORRECTLY (stale) to an IPv4 (34.224.123.113) BEFORE the reboot and the driver is sad
:
Here is the wireshark of Hubitat resolving CORRECTLY to an IPv4 (34.203.185.173) AFTER the reboot and the driver is happy
:
I would assume anything on Hubitat that is querying against a load balancer that has active changes behind it would be having a LOT of problems right now...
@gopher.ny I have a hub in the beta program (this one shown) so whenever you have a solve, I can test for you. Takes about 24hrs to see a DNS failure.


