Bug Report

We just upgraded from 3.9.2 to 3.13.0 and had saw all our pipelines failing. In our setup we rely on http(s)_proxy and no_proxy as our corporate proxy refuses to serve internal resources. Since the upgrade we can’t access internal resources anymore which means the no_proxy environment variable is not set or wrong. Inspecting running /opt/resource/check processes via /proc/$PID/environ we noticed that the no_proxy environment variable only contains the very last part the value we pass via the http_proxy environment variable to the worker: env no_proxy=host1,host2,host3 concourse worker ... ends up as no_proxy=host3 in containers scheduled by the atc to that worker.

We also confirmed that this is already wrong in the database via select name,no_proxy from workers. So the worker seems to be picking this up wrong during startup.

I opened a PR that hopefully fixes this issue: https://github.com/concourse/worker/pull/1

The following can also be handy:

• Concourse version: 3.13.0

• Deployment type (BOSH/Docker/binary): binary

• Did this used to work? Yes with 3.9.2