Skip to content

Latest commit

 

History

History
72 lines (58 loc) · 3.04 KB

platform.md

File metadata and controls

72 lines (58 loc) · 3.04 KB

Platform

Platform components are the infrastructure the lab runs on: ingress, monitoring, logging, tracing, mesh, data, secrets, security, autoscaling, cost, gitops, chaos and the dashboard. Each lives in platform/<category>/<component>/ with a single values.yaml and a chart version pinned in config/versions.env.

A target is either a category (whose provider comes from an env var, or is the only one) or an explicit category/provider for additive categories such as data.

labctl platform up                          # the full stack
MESH_PROVIDER=istio labctl platform up mesh # a category - provider from the env var
labctl platform up data/kafka               # a specific provider (additive category)
labctl platform up data/postgres
labctl platform down                        # uninstall everything
MESH_PROVIDER=istio labctl platform down mesh
labctl platform down data/kafka

platform down with no target tries every component even when one fails, then exits non-zero and lists what it could not remove. Uninstalls are safe to repeat.

labctl platform status                      # what the cluster holds, with labctl's last run
labctl platform status --live               # run each provider's status.sh
labctl platform status mesh
labctl platform status data/postgres

status lists every component the cluster holds, however it got there: platform up, scenario up --deploy-prereqs or a manual helm install. A monitoring, logging or tracing provider counts when its Helm release is deployed, since they share a namespace; any other provider counts when its namespace exists. LAST RUN is labctl's last install or uninstall of the component. When that run failed, the state is error, even if the cluster still holds what the run left behind. When the cluster does not answer, the table comes from the run history and the command says so.

status --live needs the cluster: when it does not answer, the command says so instead of printing an empty table.

Which components install is driven by the provider variables in .env: INGRESS_PROVIDER, METRICS_PROVIDER, LOGGING_PROVIDER, TRACING_PROVIDER, MESH_PROVIDER and so on. When a category has several providers and none is selected, the command lists the choices and names the variable to set. An environment variable of the same name wins over .env (which value wins).

labctl platform teardown

Removes exactly the components the inventory records as installed, in reverse order, and reports anything it could not remove instead of exiting 0 and hoping. Each uninstall is a recorded, cancellable run. The inventory holds what platform up and --deploy-prereqs installed; components installed outside labctl are not in it and are left alone.

labctl platform teardown

REST: GET /api/v2/platform, POST /api/v2/platform/{up,down}, GET /api/v2/platform/component/{category}/{name}.

Adding a component, including the CRD and StatefulSet traps that will bite you, is R05.