Outbound IPv6 requests very slow or failing to connect

Questions / Help
1

It seems that outbound IPv6 requests are either extremely slow to connect (10s+) or fail to connect entirely. I found this by debugging extremely slow outbound requests in my app and was able to isolated it via curl:

This hangs indefinitely:

curl --ipv6 https://google.com

Whereas this returns immediately:

curl --ipv4 https://google.com

I’ve verified the same issues in my Python app by logging the HTTP module and forcibly disabling ipv6.

This happens in the two regions I have my app running: yyz and syd

The only thing in my app that’s changed recently is I added multiple processes. I’m also using the latest pre-release version of litefs to support write forwarding. Seems unlikely that either of these would effect outbound TCP connections.

2

+1 from fra in May 2026. Egress to AWS eu-central-1 over IPv6 (specifically Neon’s pooler prefix 2a05:d014:d70::/48) hangs on TCP SYN with no SYN/ACK, while IPv4 to the same hosts connects immediately and curl -6 https://ipv6.google.com returns 200 — so general IPv6 egress works, but that specific prefix is unreachable. Worked around by switching uvicorn off uvloop (--loop asyncio) so a socket.getaddrinfo patch could demote AAAA for *.neon.tech.

3

Please start a new thread, this thread was from more than 3 years ago and is unrelated to the current issue in fra (and should be resolved now).