Mod fails to initialize due to slow CURL #27
Description
Hi!
Problem Description:
I recently noticed that some of my containers that should be part of my tailnet aren't, and after some investigating I found that the CURL command to download Tailscale times out after 10 minutes, and the installation of tailscale subsequently fails. I hope this is the right place to report this issue, if not I'd appreciate a pointer in the right direction!
Steps to reproduce:
The simplified version: run the curl command and a wget equivalent in a local terminal, and observe how much slower curl is:
$ curl https://pkgs.tailscale.com/stable/tailscale_1.74.1_arm64.tgz -o tailscale.tgz
$ wget https://pkgs.tailscale.com/stable/tailscale_1.74.1_arm64.tgz -O tailscale.tgz
The full repro: Run a basic nginx server through docker compose:
services:
ts-timeout:
image: linuxserver/nginx:latest
environment:
- DOCKER_MODS=ghcr.io/tailscale-dev/docker-mod:main
- TAILSCALE_STATE_DIR=/var/lib/tailscale
- TAILSCALE_HOSTNAME=ts-timeout
- TAILSCALE_AUTHKEY=tskey-auth-your-key-here
volumes:
- ts_timeout:/var/lib/tailscale
ports:
- "8083:80" # use any free port here
volumes:
ts_timeout:Run the whole thing using docker compose up. You should see something like this:
$ docker compose up ts-timeout
(... trimmed for brevity)
ts-timeout-1 | [pkg-install-init] **** Installing all mod packages ****
ts-timeout-1 | fetch http://dl-cdn.alpinelinux.org/alpine/v3.20/main/aarch64/APKINDEX.tar.gz
ts-timeout-1 | fetch http://dl-cdn.alpinelinux.org/alpine/v3.20/community/aarch64/APKINDEX.tar.gz
ts-timeout-1 | (1/4) Installing libmnl (1.0.5-r2)
ts-timeout-1 | (2/4) Installing libnftnl (1.2.6-r0)
ts-timeout-1 | (3/4) Installing libxtables (1.8.10-r3)
ts-timeout-1 | (4/4) Installing iptables (1.8.10-r3)
ts-timeout-1 | Executing busybox-1.36.1-r29.trigger
ts-timeout-1 | OK: 167 MiB in 191 packages
ts-timeout-1 | /run/s6-rc:s6-rc-init:BALEpN/servicedirs/s6rc-oneshot-runner /run/s6-rc:s6-rc-init:BALEpN/servicedirs/s6rc-oneshot-runner
ts-timeout-1 | % Total % Received % Xferd Average Speed Time Time Time Current
ts-timeout-1 | Dload Upload Total Spent Left Speed
100 3308 100 3308 0 0 2910 0 0:00:01 0:00:01 --:--:-- 2914
ts-timeout-1 | % Total % Received % Xferd Average Speed Time Time Time Current
ts-timeout-1 | Dload Upload Total Spent Left Speed
(...nothing happens for a long time...)
ts-timeout-1 | curl: (92) HTTP/2 stream 1 was not closed cleanly: INTERNAL_ERROR (err 2)
ts-timeout-1 | s6-rc: warning: unable to start service init-mod-universal-tailscale-postinstall: command exited 92
The first request gets the current version and finishes quickly, the second one is the one to download the tarball, and it'll fail after 10 minutes.
Running the install script manually works (but that doesn't help much since I'd have to manually run it every time the container restarts).