Getting connection reset on 53/tcp

1

Still testing my DNS service and UDP query responses have been 100% for the past week which is great.

However, I’ve just tested DNS TCP and unfortunately it no longer seems to work. E.g.,:

$ dig @213.188.193.79 wombatsoftware.com +tcp
;; communications error to 213.188.193.79#53: connection reset

My DNS server is running both 53/udp and 53/tcp and they’re both configured in my fly.toml. App name is slickdns.

Weirdly the 53/tcp port is configured as the health check and all instances are showing healthy…

2

@jbarham Sorry about this. Looks like a change we made had unforeseen consequences for port 53. We reverted the commit after much debugging and it appears to be working properly again.

3

Fix confirmed. Thanks!

4

This is mostly my fault, for what it’s worth; we built an off-network infrastructure for monitoring UDP DNS a few weeks ago, and I should have done TCP at the same time (UDP is routed differently than TCP in our network and so I had a lot of specific UDP things I wanted to monitor, and it got away from me.)

We should end the weekend with monitoring that pages me out of bed if 53/tcp starts getting routed weirdly.