Container status API: compute live status instead of static DB column

[runleveldev]

Resolves #350. Merge #351 first.

Summary

Container status was a static Containers.status column updated only by the create/reconfigure job scripts, so it drifted from reality whenever Proxmox changed out of band or a job died mid-run. This replaces it with status computed live on demand, embedded directly on the existing container API payloads. The static column is dropped.

Live status resolver — create-a-container/utils/container-status.js

Status is derived from two cheap sources of truth:

• Proxmox run-state — does the LXC exist and is it running/stopped? Read from a single clusterResources('lxc') snapshot per node.
• Create job — for containers not yet in Proxmox: is the create job active, did it fail, or is it gone? The create job is linked by the creationJobId foreign key, so this is resolved from the eager-loaded creationJob association (zero extra queries) or a single PK lookup.

Resolved statuses:

Status	Condition
running	online in Proxmox
offline	exists in Proxmox but stopped
creating	no Proxmox VM yet, active create job
failed	no Proxmox VM, create job returned failure
missing	no Proxmox VM, create succeeded or no create job found
unknown	Proxmox unreachable / node has no API credentials

Where status is returned

The live status is embedded on the existing list, show, and create container payloads — no new endpoint.

Earlier revisions of this PR also implemented restarting (active reconfigure job) and out-of-sync (live LXC config differs from expectation), plus a dedicated GET /containers/:id/status endpoint. All three were removed for performance/simplicity (see below). The PUT endpoint still enqueues a reconfigure job and returns its jobId; only the restarting status label is gone.

Performance (why the scope was trimmed)

Resolving the list must stay cheap. Two things were inherently O(N):

• Reconfigure-job lookups (for restarting) — reconfigure jobs have no FK, so detecting them needed per-container command-string queries.
• Config-drift reads (for out-of-sync) — Proxmox has no bulk config endpoint, so this needed one lxcConfig call per in-Proxmox container.

Both were dropped. The index now issues one Proxmox call per node and zero per-container Proxmox or DB job queries (verified at 30 containers). The create job comes free from the eager-loaded association.

Compatibility

• status remains present on the same endpoints, so existing consumers keep working. Checked mieweb/launchpad (an external API client): it only checks status == "running" and never uses restarting/out-of-sync, so dropping them is safe there.
• This narrows the status set vs. what main could emit (main could return restarting), so it's not a strict superset anymore — but no known client depends on the removed values.

Removed the static column

• Migration 20260615000000-remove-container-status.js drops Containers.status; removed from the model.
• All former writers updated to stop persisting status (create-container.js, reconfigure-container.js, nodes.js import, json-to-sql.js). Failure is now captured by the job's failure status.
• The create-time status !== 'pending' guard became a Proxmox-VMID-presence guard.
• routers/templates.js (nginx/dnsmasq config) previously filtered where: { status: 'running' }; it now selects routable containers by ipv4Address != null.
• routers/api/v1/resource-requests.js (applyResourceToExistingContainers) previously filtered where: { ..., status: 'running' } — that would have thrown "Unknown column 'status'" after the migration. It now filters on containerId != null (a VMID is required to reconfigure anyway).

Frontend

• Typed ContainerStatus union; the list page renders the live status from the list response via a small presentational StatusBadge (friendly labels + status → badge-variant mapping). No per-row polling.

Docs

• OpenAPI: ContainerStatus enum, referenced from the Container schema.
• README: ER diagram (removed status) and a container-status field description.

Testing

No test framework exists in the repo, so logic was validated with standalone Node assertion scripts (not committed):

• Decision-table assertions for every status path.
• Batch-resolution assertions confirming 0 per-container Proxmox/DB calls and 1 cluster snapshot per node at scale (30 containers).
• All backend files pass node --check; OpenAPI YAML parses; client type-check and build pass.

Notes / follow-ups

• Still a draft. This drops a DB column, so coordinate with deploy/migration timing.
• Config-drift (out-of-sync) can be revisited later if Proxmox gains a bulk config endpoint or if drift is computed lazily on the detail view / cached.
• Happy to add a prove/test harness if the team wants committed tests.

[Copilot]

Pull request overview

This PR removes the persisted Containers.status column and replaces it with a live, computed container status derived from Proxmox state, job state, and config drift, while keeping status present on existing container API payloads. It also updates the UI and docs to reflect the expanded status set.

Changes:

• Add a shared status resolver (utils/container-status.js) and wire it into container list/show/create responses.
• Drop Containers.status from the Sequelize model + DB schema; remove status writes from job scripts/import paths.
• Update templates generation filters and frontend types/rendering to support the new live status values.

Reviewed changes

Copilot reviewed 10 out of 13 changed files in this pull request and generated 6 comments.

Show a summary per file

File	Description
create-a-container/utils/container-status.js	New live status resolver (Proxmox snapshot + job lookup + drift detection).
create-a-container/routers/templates.js	Switch template container selection from status='running' to a provisioned/proxy predicate.
create-a-container/routers/api/v1/nodes.js	Stop persisting status during node/container import.
create-a-container/routers/api/v1/containers.js	Embed computed status in list/show; return creating on create response; stop persisting status updates.
create-a-container/README.md	Remove status from ER diagram and document the live status field and enum values.
create-a-container/openapi.v1.yaml	Add ContainerStatus enum schema and reference it from Container.status.
create-a-container/models/container.js	Remove status attribute from the Container model definition.
create-a-container/migrations/20260615000000-remove-container-status.js	Migration to drop the status column (with rollback restoring a default).
create-a-container/client/src/pages/containers/ContainersListPage.tsx	Render typed live status via StatusBadge and map new states to badge variants/labels.
create-a-container/client/src/lib/types.ts	Introduce ContainerStatus union and type container/status-bearing payloads accordingly.
create-a-container/bin/reconfigure-container.js	Stop updating container status; keep MAC/IP updates and adjust logs.
create-a-container/bin/json-to-sql.js	Stop setting/updating container status during JSON import/update.
create-a-container/bin/create-container.js	Replace the old “pending” guard with a VMID presence guard; stop persisting creating/running/failed status.