Benchmark results aren’t fetched from Cachix about primer HOT 3 CLOSED

Hmm, looks like this has been happening for awhile; e.g.:

https://github.com/hackworthltd/primer/actions/runs/8592173489/job/23541841064#step:6:129

We’re about to switch to tasty-bench, so between that and having just switched benchmark hardware, maybe this is a good time to throw out the old results and start fresh.

I’m not sure why the benchmark results aren’t being cached, though. I’ll compare the hashes of the Nix derivaitons (Buildkite vs GitHub) to see if there’s an impurity somewhere. I’ll also ensure that cachix watch-store is being run on the benchmarking machine, though perhaps we should go back to Nix-style benchmarking by adding the new benchmarking machine to /etc/nix/machines in the Buildkite pod on our Kubernetes cluster, which would ensure that the result is cached since the Buildkite pod is definitely writing to Cachix.

from primer.

The Nix hashes are the same, which indicates that the results aren’t being written to Cachix… but it might be the case that we’re writing the benchmark results to our private Cachix cache and not the public cache that the GitHub workflow relies on. (Nope, scratch that bit — the benchmark GitHub workflow does have access to the private Cachix, and in fact you can see it downloading some Nix derivations from there during the workflow.)

from primer.

This has been fixed.

from primer.