In default mode a layer exist only once in the cache, and regardless
which repository it is initially from, each repository get a hit on
this layer now.
Pitfall: You can not push images that reuse layers from a different
registry (`ALLOW_PUSH=true`).
Setting this option to 'false' adds the host to the cache key to solve this.
Default is `true` to not change default behavior.
WARNING: Changing this setting invalidates your cache!
If push is allowed, buffering requests can cause issues on slow upstreams.
The put requests pile up, and since `proxy_ignore_client_abort` is on...
...the push fails.
If you have trouble pushing, set `PROXY_REQUEST_BUFFERING`to false, then fix
remainig issues/timeouts.
Default is true to not change default behavior.
* implement manifest caching; refactor config with includes, and generate from ENVs in entrypoint.sh
- disabled by default; enable with -e ENABLE_MANIFEST_CACHE=true
- default times and regexes are a wild guess, make sure to tune for your use case.
- add manifest caching/anti-ratelimit usage note to README
- add -e ENABLE_MANIFEST_CACHE=true to examples, some wording changes
- add -e ENABLE_MANIFEST_CACHE=true to one the steps in test workflow.
- injects a mitmproxy between the caching layer and DockerHub's registry-1.docker.io
- it also disables VERIFY_SSL so the mitmproxy can be used;
- DEBUG_HUB's mitmweb is exposed on port 8082
- both DEBUG and DEBUG_HUB can be enabled independently
- now only /v2/.../blobs/... URIs are actually cached (together with their redirect catchers)
- /manifests/, /token, and /v2/ are not cached anymore, which should solve a lot of problems
- better messages for /v1 attempts
- fix usage of $connect_host:443 (which is hostname:port and causes errors to be logged) to $connect_addr (which returns an IP:port) in the proxy layer
saces
Ricardo Pardini
Sezer Bülent
Federico Gimenez
ricardop
Jeff Bachtel