Currently, the rock testing workflow depends on microk8s registry, skopeo copying the rock to the microk8s local registry, and using kgoss.
Recently we started seeing timeout failures.
|
sudo concierge prepare -p microk8s --extra-snaps rockcraft,just,kubectl |
However,
- We now use "k8s" rather than "microk8s", so testing this locally means deploying a dedicated VM with
conierge prepare -p microk8s instead of -p dev.
- The current approach has many dependencies which are fragmented across this repo and the rock's justfile.
- Use of kgoss with the microk8s registry is yet another level of indirection which complicates things.
We can accomplish the same with a single dependency - containerd - regardless of which k8s is installed, for example:
sudo ctr image import --base-name grafana/agent grafana-agent_0.44.6_amd64.rock
sudo ctr run --net-host -d grafana/agent:0.44.6 ga
# Run regular goss here
curl localhost:12345/agent/api/v1/targets
sudo ctr task kill ga
sudo ctr container rm ga
Currently, the rock testing workflow depends on microk8s registry, skopeo copying the rock to the microk8s local registry, and using kgoss.
Recently we started seeing timeout failures.
observability/.github/workflows/rock-pull-request.yaml
Line 43 in 78f655b
However,
conierge prepare -p microk8sinstead of-p dev.We can accomplish the same with a single dependency - containerd - regardless of which k8s is installed, for example: