Proxy-Cache for official images isn't working

[fandujar]

Expected behavior and actual behavior:
Commands like docker pull myharbor.mycompany.com/myproxycache/alpine:latest would result in image pulled

Steps to reproduce the problem:

• Create a project with proxy-cache enabled linked to dockerhub endpoint (i used dockerhub as my project name)
• Use docker client to pull a official image like alpine or nginx

Versions:
Please specify the versions of following systems.

• harbor version: [2.1.1]
• docker engine version: [19.03.13 (Community)]

Result from docker pull command:

docker pull myharbor.mycompany.com/dockerhub/alpine:latest Error response from daemon: unknown: repository dockerhub/alpine not found

[EdKingscote]

Have you tried docker pull myharbor.mycompany.com/dockerhub/library/alpine:latest there is a footnote at the bottom of the docs at https://goharbor.io/docs/2.1.0/administration/configure-proxy-cache/#how-harbor-proxy-cache-works about single level repositories and official images.

[alfsch]

I have the same errors with exact the same harbor setup. When I use

docker pull myharbor.mycompany.comdockerhub/library/alpine:latest

I get a different error
Error response from daemon: missing or empty Content-Type header

When I use the command like in the bug description I get the same error as in issue description. The used docker version is:

Docker version 20.10.0, build 7287ab3

[fabianotessarolo]

Yeah, indeed, that worked here @EdKingscote @fandujar.

A related issue here, but for Transparent proxy is:
Problem: #8082
Feature Request: #13579

Seems like NGINX rewrite suggestion for Transparent Proxy, should take into consideration this.

[fandujar]

@EdKingscote I tried, but got same error @alfsch

Error response from daemon: missing or empty Content-Type header

[alfsch]

@fandujar and @EdKingscote: the problem is docker 20.10 see #13740

[EdKingscote]

I ran in to the Content-Type header problem last night as well, but only on machines that all just happened to be internal, so I figured it was a quirk of HAProxy being our front-end before I had to call it a night. Thanks @alfsch for sharing the link to that new issue, as it turns out purely by chance all of our external hosts are using older docker engines than our internal ones - that saved me a good chunk of investigation time this morning! Hopefully it won't take too long for a fix to materialize somewhere.

[fandujar]

Tks for sharing and for the help @EdKingscote @alfsch . I will close this one then.