immich-app/yucca-o11y
1
0
Fork 0
mirror of https://github.com/immich-app/yucca-o11y.git synced 2026-09-30 13:23:23 +08:00
Code Issues Packages Projects Releases Wiki Activity

docs(o11y): say what stays stable across publishes (#326)

Browse Source 
Signed-off-by: Devin Buhl <devin@buhl.casa>
This commit is contained in:
Devin Buhl
2026-09-13 09:03:22 -04:00
committed by GitHub
parent 9396a2dd72
commit 3d1385017c
2 changed files with 2 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files
.mise/config.toml
+1 -1
View File
@@ -42,7 +42,7 @@ sync_job = "docker run --rm -e OUTPUT=- -e CONFIG=/config/sync-job.yaml -e NAMES
# on the default (tenant-scoped) datasource and show only this cluster.
pin_fleet = "yq '(select(.kind == \"GrafanaDashboard\") | .spec.json) |= (fromjson | .templating.list[] |= (select(.type == \"datasource\") |= (.regex = \"/^VictoriaMetrics Fleet$/\" | .current = {\"text\": \"VictoriaMetrics Fleet\", \"value\": \"VictoriaMetricsFleet\"})) | tojson)'"
# The sync-job emits the kube-prometheus bundle in map order; sort documents so an unchanged
# upstream renders byte-identical and the published artifact keeps its digest.
# upstream renders byte-identical and the artifact's content layer keeps its digest.
sort_docs = "yq eval-all '[.] | sort_by(.metadata.name) | .[] | splitDoc'"
tg = "op run {% if get_env(name='OP_SERVICE_ACCOUNT_TOKEN', default='') == '' %}--account 'team-futo.1password.com' {% endif %}'--env-file={{config_root}}/deployment/.env' -- terragrunt"
o11y/README.md
+1 -1
View File
@@ -12,7 +12,7 @@ mise run //:o11y:render # render manifests/dashboards.yaml locally for review
Both tasks first run `o11y:vendor`, which fetches boards that need a rewrite the sync-job cannot express and writes them to `vendor/` as local sources. Today that is CloudNativePG only: upstream uses `cluster` to mean the Postgres cluster, while the fleet keeps that name under `pg_cluster` and reserves `cluster` for the Kubernetes cluster, so the vendor step renames the label and variable before the sync-job adds the fleet's `$cluster`. A cluster's CNPG series must carry `pg_cluster` for the board to list its databases; the shipping guide's identity-label section shows the relabel rule, which o11y, azad and harbor apply.
The set is everything at least two clusters run: the dotdc Kubernetes views and system boards, the kube-prometheus mixin boards (kubelet, scheduler, controller manager, proxy, API server, compute resources, networking, persistent volumes, node exporter USE method), Node Exporter Full, etcd, vmagent, Cilium and Hubble, Flux, Spegel, Envoy Gateway and CloudNativePG. Windows, AIX, macOS, Prometheus, Alertmanager and Grafana-overview boards from the mixin bundle are disabled. Documents are sorted by name so an unchanged upstream renders byte-identical and the published artifact keeps its digest.
The set is everything at least two clusters run: the dotdc Kubernetes views and system boards, the kube-prometheus mixin boards (kubelet, scheduler, controller manager, proxy, API server, compute resources, networking, persistent volumes, node exporter USE method), Node Exporter Full, etcd, vmagent, Cilium and Hubble, Flux, Spegel, Envoy Gateway and CloudNativePG. Windows, AIX, macOS, Prometheus, Alertmanager and Grafana-overview boards from the mixin bundle are disabled. Documents are sorted by name so an unchanged upstream renders byte-identical: the artifact's content layer keeps its digest across publishes, and only the manifest's revision and timestamp annotations change, so Flux's re-apply after a publish with no upstream change is a no-op.
Upstream sources are pinned where the upstream moves (Cilium by release tag, Flux by commit) and tracked at `master` where the VictoriaMetrics chart does the same.