Plan and deploy

Planning evaluates the complete template pipeline without provisioning resources. Deployment submits the evaluated graph to the durable orchestrator.

Preview changes

praxis plan stack.cue \
  --account prod \
  -f prod.vars.json

The plan compares desired resources with stored Praxis state and live provider observations. It reports creates, updates, replacements, deletes, and unchanged resources with field-level differences when the driver exposes them.

Save a plan for review:

praxis plan stack.cue \
  --account prod \
  -f prod.vars.json \
  --out payments.plan.json

Apply a reviewed plan

praxis deploy \
  --plan payments.plan.json \
  --key payments-prod \
  --wait

Praxis checks the saved plan’s content hash and age before applying it. Signed plans can add an integrity boundary for a review pipeline.

Useful controls

# Plan only one resource and its dependencies.
praxis plan stack.cue --account prod --target application

# Force an explicit replacement.
praxis deploy stack.cue --account prod --replace application

# Allow replacement when immutable fields changed.
praxis deploy stack.cue --account prod --allow-replace

# Return immediately after submission.
praxis deploy stack.cue --account prod --key payments-prod

Independent graph nodes run concurrently. --parallelism can impose a deployment-wide limit when provider quotas or change-management policy require it.

Failure behavior

A failed resource stops its dependents and leaves independent completed resources in their known state. Praxis does not automatically roll back successful resources after every deployment failure. Operators can inspect the causal chain and choose deletion, a corrected apply, or a point-in-time rollback.

Full reference

For deployment scheduling, durable state, failure propagation, rollback, and recovery, see Orchestration in the GitHub documentation. For every command option, see the CLI reference.