When using Jenkins behind an HTTP proxy, the EC2 launcher also tries to route traffic through the proxy.

To reproduce

Configure an HTTP proxy in Manage Jenkins > Manage Plugins > Advanced
Set a Server and Port for an HTTP Proxy.
Then configure the EC2 plugin and try to launch a slave.

What I expect

The HTTP proxy, which is there for jobs that need to connect by HTTP to the Internet ref, should be ignored, because the EC2 plugin is not trying to open an HTTP connection, nor is it normally trying to send traffic to the Internet.

What actually happens

An error message might be seen in the slave instance launch log like:

INFO: Failed to connect via ssh: HTTP Proxy Error (403 Forbidden)

Eventually, the slave will fail to launch.

More info

It looks like this issue was introduced in JENKINS-18855 in this commit. Commit message notes, "It uses HTTP proxy defined in Plugin configuration. SOCKS proxy is not supported".

This issue was reported again e.g. in JENKINS-30041 and a workaround was proposed, namely adding no_proxy=10.x.x.x. This issue was then closed, however, and the workaround there might work sometimes but not always.

A better solution would be to add a SOCKS proxy feature for the EC2 plugin rather than inheriting proxy settings from Jenkins.