Failing to deploy using Github Actions

I am having issues deploying using github actions. Have read through this thread it seems Github actions create a WireGuard peer connection per deploy.

Having listed all my peed connections with fly wireguard list shows I have 18. I have tried deleting them but am getting errors.

Removing WireGuard peer "interactive-agent-fv-az203-563-services-eriade-com-01GC906PE2W6PMZ3Z8RJNHQW30" for organization eriade-410 Error Post "": http2: server sent GOAWAY and closed the connection; LastStreamID=2147483647, ErrCode=NO_ERROR, debug=""

The only way I seem to be able to deploy is to delete by builder and then redeploy.

What is the correct way of fixing this once and for all? I do not want to keep deleting my builder every time. Thanks

Haven’t ran in to this issue myself, but does resetting wg connection (from within gh-action) not work before executing deploy?

➜  flyctl wireguard reset <org-name>
New WireGuard peer for organization '<org-name>': 'interactive-agent-<user-identity>-RANDOMBASE32'

I don’t think one could do much other than complain, but I expect Fly to sort this out, given that this issue is inherent in their software and architecture.

I stated that “I have tried deleting them but am getting errors.”

I do not know who you are refering to as “complaining” but that is not the reason I created this Topic. It might be related but I do not know. My error may be different and could help Fly resolve the issue quicker instead of going on the other thread with “me too”. Only to find out that my error was completely unrelated…

Thank you for your response nevertheless.

I believe you may have misjudged the tone of my post.


wg reset does not delete anything, but in fact creates a new identity. A slightly different command. Though, I have no clue if it resolves the issues you’re seeing (I haven’t encountered them).

Gotcha. But, what I meant is: There isn’t much we, as users, could do (other than workaround this issue and complain about it). Thomas, Principal at Fly, in the other thread, asked to users “please complain”, 'cause it sounds like it is their current design which is at its limit here and their short-term solutions may stop working any given minute.