Update flake lock

[conscious-puppet]

No description provided.

[srid]

What's the context for this change? Is services-flake not supported on nixpkgs older to this update anymore?

[conscious-puppet]

in the PR #115, i made a mistake with the setup command for pgadmin 8.2. since the older nixpkgs had pgadmin 7.5, i could not test this and it got missed.

for older nixpkgs, pgadmin has pgadmin4-setup executable to setup the configuration db, and for newer nixpkgs, it has pgadmin4-cli executable to setup the db. so services-flake would support both

  if [ -f ${config.package}/bin/.pgadmin4-setup-wrapped ]; then
    # pgadmin-7.5 has .pgadmin4-setup-wrapped
    ${config.package}/bin/.pgadmin4-setup-wrapped setup
  else
    # pgadmin-8.2 has .pgadmin4-cli-wrapped
    ${config.package}/bin/.pgadmin4-cli-wrapped setup-db
  fi

[shivaraj-bh]

Let’s understand if we really need this: #117 (comment)

Since NixOS/nixpkgs#283330, along with the go backend, grafana frontend is also built from source in nixpkgs. But the offline cache doesn’t support darwin, this can be fixed by overriding it:

{
  grafana = prev.grafana.overrideAttrs (oa: rec {
    offlineCache = oa.offlineCache.overrideAttrs (_ :{
      buildPhase = ''
        runHook preBuild
        export HOME="$(mktemp -d)"
        yarn config set enableTelemetry 0
        yarn config set cacheFolder $out
        yarn config set --json supportedArchitectures.os '[ “linux” “darwin" ]'
        yarn config set --json supportedArchitectures.cpu '[ “x64” “arm64" ]'
        yarn
        runHook postBuild
      '';
      outputHash = lib.fakeHash;
    });
  });
}

Note: replace the outputHash with whatever the offlinecache generates.

[shivaraj-bh]

Even after fixing the frontend cache, the go backend fails to build. There is an upstream issue open: NixOS/nixpkgs#273998

Looks like it is do with nix: NixOS/nix#9977

[shivaraj-bh]

As a workaround, we can use grafana < 10.2.0 until the case-insensitivity issue is resolved.

[shivaraj-bh]

Turns out we can get the latest grafana build on darwin, thanks to: NixOS/nixpkgs#129730

[shivaraj-bh]

CI fails with: no space left on device 🙂

[shivaraj-bh]

Until we self-host the runner, hope this helps: actions/runner-images#2875

[srid]

@conscious-puppet @shivaraj-bh What's the original reason for updating flake.lock in this PR? If we revert this, can we avoid having to build grafana on CI and locally?

[conscious-puppet]

some of the packages were old.

• the default package for cassandra was cassandra 3. which had a dependency on python 2, and was no longer supported on aarch64-darwin
• pgadmin4 also had an outdated version with different cli for setup db.

if we revert this, then we may not be able to test services flake against current package versions with just run <service> or just test <service>.

[conscious-puppet]

if build is an issue, can we overlay grafana to an older version for darwin?

[srid]

can we overlay grafana to an older version for darwin?

This is what I'm trying now