Does changing the address of a *dagger.Socket on invalidates the cache

[chilly vault]

I have a pipeline running in an azure devops agent where the value of SSH_AUTH_SOCK changes dynamically on every build.

My build pipeline relies on that socket for SSH Authentication before the build of my code (to retrieve dependencies).

This is the pseudo code for my pipeline :

- New(*dagger.Socket)
- Build()
  - Set SSH Socket + SSH_AUTH_SOCK env variable on container
  - Retrieve dependencies (git pull etc )
  - Build the code

I noticed that i'm not hitting the cache. Is that because the address of the socket changes ? If so is there another way ?

Does changing the address of a *dagger.Socket on invalidates the cache

[grand falcon]

yes, it will invalidate the cache since it will re-download the dependencies as the SSH socket path changed. One thing you can do to prevent this is to create a symlink always in the same path before running dagger call and this will prevent the cache invaliation you're seeing

[chilly vault]

Thanks @grand falcon ! I will give it a shot and report back 🙂

[chilly vault]

@grand falcon you legend. It indeed did the trick

- bash: |
    set -euo pipefail
    mkdir -p "$HOME/.ssh"

    ln -sf "$SSH_AUTH_SOCK" "$HOME/.ssh/agent-socket"

    echo "##vso[task.setvariable variable=SSH_AUTH_SOCK]$HOME/.ssh/agent-socket"
    echo "SSH_AUTH_SOCK is now: $SSH_AUTH_SOCK"

[grand falcon]

Glad that worked out 👐

[chilly vault]

down from 13min to 25sec when no code changes

[grand falcon]

sweet. Is 25s with the entire pipeline cached? or does it effectively run something? Mostly asking since 25s for a pipeline which is fully cached seems a bit over what I'd expect 🙏

[chilly vault]

in this case it didnt run anything

[grand falcon]

do you have a trace of that pipeline by any chance?