observability: implement better/fully working grafana sync (aptos-lab…

…s#5064)

The previous grafana sync method had a few bugs and didn't sync/support certain constructs and led broken dashboards - this was likely due to the Go grafana-sync tool using a kinda old Grafana SDK version and relying on Go structs having implemented every single possible field.

With the new sync method we're using the API directly and treating the dashboard more-or-less as a blob so it should work mostly fine, even when grafana adds new properties/features in the serialization format.

Also adding a method to sync dashboards back from local to Grafana which wasn't really supported before.

Lastly this also auto deletes dashboards that don't exist in Grafana anymore.

.github/workflows/grafana-sync.yaml

@@ -18,8 +18,14 @@ jobs:
     runs-on: ubuntu-latest
     steps:
       - uses: actions/checkout@93ea575cb5d8a053eaa0ac8fa3b40d7e05a33cc8 # pin@v3
-      - name: Pull and sync Grafana dashboards
-        run: ./scripts/sync_grafana_dashboards.sh
+      - uses: pnpm/action-setup@537643d491d20c2712d11533497cb47b2d0eb9d5 # pin https://github.com/pnpm/action-setup/releases/tag/v2.2.3
+        with:
+          version: 7.13.4
+          run_install: |
+            - recursive: false
+              args: [--frozen-lockfile]
+      - name: Download current Grafana dashboards
+        run: pnpm exec scripts/grafana-sync.mts download
       - name: Create Pull Request
         uses: peter-evans/create-pull-request@671dc9c9e0c2d73f07fa45a3eb0220e1622f0c5f # pin@v4
         with:

.gitignore

@@ -98,8 +98,5 @@ node_modules
 # Cached python files
 __pycache__
 
-# Grafana sync
-grafana-sync*
-
 # ignore framework release bundle file
 **/*.mrb