Unable to start any Docker containers that bind ports

[ljmccarthy]

I have some scripts that update and restart my long-lived containers and recently they stopped working.

For example, I am receiving the following error message:

docker: Error response from daemon: driver failed programming external connectivity on endpoint syncthing (409740eeab268768a0071d8d034403a24a11a8399ad351aa3fb2a85928b82fe5): Error starting userland proxy: listen tcp [::]:22000: bind: address already in use.

I have verified with lsof that the ports are not in use by any other process.

Note: I am using the runc runtime, not Kata containers.

[ljmccarthy]

Here's a trivial example which can be used as a test case:

ljmcc@pythagoras~ $ docker run -t -i --rm ubuntu bash
Unable to find image 'ubuntu:latest' locally
latest: Pulling from library/ubuntu
3153aa388d02: Pull complete 
Digest: sha256:0bced47fffa3361afa981854fcabcd4577cd43cebbb808cea2b1f33a3dd7f508
Status: Downloaded newer image for ubuntu:latest
root@2e3c52d6fe66:/# 
exit
ljmcc@pythagoras~ $ sudo lsof -i :12345
ljmcc@pythagoras~ $ docker run -p 12345:12345 -t -i --rm ubuntu bash
docker: Error response from daemon: driver failed programming external connectivity on endpoint gallant_antonelli (06282f5b0effbe7b8c0a15bf0f67f33bc02f38c94e77711eecc48b5e199e7eec): Error starting userland proxy: listen tcp [::]:12345: bind: address already in use.

[bryteise]

See https://github.com/clearlinux/distribution/issues/2918

[apwiggins]

What's the source for upstream change? Is this an actual directed change or an upstream reportable bug? Docker release notes for v24 make no mention of changes to the ports specfication -- which is a pretty signifcant change given that it would break most docker-compose YAMLs and bash scripts. https://docs.docker.com/engine/release-notes/24.0/

Docker documentation (which I'm assuming is current) shows that a plain port specification is valid. It only mentions using an IP address as an optional improvement https://docs.docker.com/network/ The Clear Linux forum post and workaround doesn't really clear the air and doesn't show that upstream changed port specification behaviour intentionally. This is still a problem and seems to deviate from upstream Docker's intentions.

unfortunately upstream docker changed their syntax for this see also https://community.clearlinux.org/t/docker-issue-on-clear-linux/8927 for discussion/solutions … On Sun, Jul 2, 2023 at 11:08 AM moroznah @.> wrote: Multiple VMs updated to 39540, Docker version 24.0.2 no longer starts any containers with port mapping ("-p" option). Ex. error: ERROR: for webapp Cannot start service webapp: driver failed programming external connectivity on endpoint webapp (8b4185c591f7adeba74997fdc78b78bff436dc5cb3bb8efd5b80a4b2f33f8982): Error starting userland proxy: listen tcp [::]:8000: bind: address already in use All these ports are not used and multiple containers worked prior to update. — Reply to this email directly, view it on GitHub https://github.com/clearlinux/distribution/issues/2918, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAJ54FMCBL7Y77ED46C6NLDXOG2J7ANCNFSM6AAAAAAZ3SUWJU . You are receiving this because you are subscribed to this thread.Message ID: @.>