Temporal variants

OpenBox uses Temporal as its workflow engine. The Helm chart supports two variants — customer picks per data residency, ops burden, and cost profile.

Side-by-side comparison

Aspect	Self-hosted Temporal	Temporal Cloud
Workflow engine	4 in-cluster pods (frontend / history / matching / worker) — image `temporalio/server:1.29.1`	Managed SaaS
Client → engine protocol	gRPC plaintext, cluster-internal `temporal-frontend.temporal.svc:7233`	gRPC + TLS across public internet
Visibility (search) store	Elasticsearch 7.17.3 × 3 pods (30 GiB gp3 each)	Temporal-managed
Default (workflow history) store	Dedicated PostgreSQL cluster (`openbox-postgres-temporal`)	Temporal-managed
Temporal Web UI	Self-hosted `temporal-web` pod behind Istio	Temporal SaaS console
Ops burden	HIGH — cluster upgrades, ES health, retention tuning, PG backup + DR	LOW — Temporal handles upgrades + retention
Data residency	Fully in your VPC	Temporal SaaS region (limited list)
AWS bill delta	Baseline + ~$300-640/mo (memory pool + DB pool + ES + PVCs)	Baseline
Temporal Cloud subscription	$0	~$50-500+/mo (workflow-volume dependent)

Break-even math

Roughly: self-hosted Temporal wins on total cost if Temporal Cloud's per-action billing exceeds ~$300-600/mo for your workload. Below that threshold, Temporal Cloud is cheaper (no AWS delta + minimal SaaS bill).

Rule of thumb:

< 100K workflow executions/day → Temporal Cloud
> 500K workflow executions/day → Self-hosted starts to win
Between: compute both, pick whichever has more headroom

Openbox does not prescribe — customer knows their workload.

When to pick self-hosted

✅ Data residency requirement forbids SaaS (finance, gov, EU-strict)
✅ Air-gapped or heavily restricted network (S4 post-MVP scenario)
✅ Very high workflow volume (self-hosted amortizes fixed infra cost)
✅ Existing Temporal expertise on team (upgrades, retention, ES ops)

When to pick Temporal Cloud

✅ Small-to-medium workflow volume
✅ Team doesn't want ES / PostgreSQL ops on top of application ops
✅ Multi-region setup — Temporal Cloud handles regional replication
✅ Fast time-to-first-workflow (no in-cluster Temporal to provision)

Chart config

Set via openbox-core.temporal.mode:

# Self-hosted variant
openbox-core:
  temporal:
    mode: self-hosted
    server:
      replicas: 4

# Temporal Cloud variant
openbox-core:
  temporal:
    mode: cloud
    cloud:
      namespace: <your-temporal-cloud-namespace>
      address: <ns>.<acct>.tmprl.cloud:7233
      tlsSecretRef: temporal-cloud-tls    # K8s Secret with client cert + key

Full values reference: openbox-core.temporal.*.

Migration between variants

Cloud → Self-hosted (or vice versa) requires:

Drain in-flight workflows (Temporal API: SignalWithStart to a terminate flow)
Export workflow history (self-hosted: pg_dump; Cloud: Temporal export API)
Import to target variant
Cut over openbox-core.temporal.* values + helm upgrade
Verify workflow visibility in target UI

Zero-downtime migration is not supported by the chart. Plan a maintenance window (~1-4 hours depending on workflow volume).

Side-by-side comparison​

Break-even math​

When to pick self-hosted​

When to pick Temporal Cloud​

Chart config​

Migration between variants​