Checkpoint configuration

Checkpoints enable durable execution by saving state at defined points, allowing missions to resume after failures.

Basic usage

Add a checkpoint configuration at the mission level:

mission DurableSync {
  checkpoint: afterStep

  action FetchData {
    get "/items"
    store response -> items { key: .id }
  }

  run FetchData
}

Checkpoint modes

afterStep

Saves state after every step completes successfully:

mission CriticalPipeline {
  checkpoint: afterStep

  action Process {
    get "/data"           // Checkpoint saved after this
    store response -> data // Checkpoint saved after this
    for item in data {
      post "/process"     // Checkpoint saved after each iteration
    }
  }

  run Process
}

Best for: Maximum durability, critical workloads, expensive operations.

onFailure

Only saves state when a step fails:

mission EfficientSync {
  checkpoint: onFailure

  action Sync {
    get "/items"
    store response -> items { key: .id }
  }

  run Sync
}

Best for: Better performance when failures are rare, less critical workloads.

How checkpoints work

State captured

Each checkpoint captures:

Component	Description
Stage index	Current pipeline stage
Step index	Current step within the action
Variables	All variable values at that point
Response	Current response value
Store state	References to store contents
Loop context	Iterator position if in a loop

Resume behavior

When resuming from a checkpoint:

1. Reqon loads the last checkpoint
2. Execution starts from the next step after the checkpoint
3. Variables and context are restored
4. Stores are reconnected (data persists separately)

# Resume an interrupted mission
reqon mission.vague --resume

Checkpoint storage

Default storage

Checkpoints are stored in .vague-data/execution/:

.vague-data/execution/
├── SyncMission-2024-01-20T09-00-00.json
├── SyncMission-2024-01-20T10-00-00.json
└── checkpoints/
    └── SyncMission-exec-abc123.json

Checkpoint file structure

{
  "executionId": "exec-abc123",
  "mission": "SyncMission",
  "timestamp": "2024-01-20T09:15:30Z",
  "stageIndex": 1,
  "stepIndex": 3,
  "action": "ProcessData",
  "variables": {
    "page": 5,
    "total": 100
  },
  "response": { "items": [...] },
  "loopContext": {
    "itemIndex": 42,
    "collection": "items"
  }
}

Use cases

Large dataset processing

mission ProcessLargeDataset {
  checkpoint: afterStep

  action FetchAll {
    get "/items" {
      paginate: offset(page, 1000),
      until: length(response) == 0
    }
    store response -> items { key: .id }
  }

  action ProcessAll {
    for item in items {
      post "/process" { body: item }
    }
  }

  run FetchAll then ProcessAll
}

If processing fails at item 5000, resume picks up from item 5000.

Multi-stage pipelines

mission ETLPipeline {
  checkpoint: afterStep

  action Extract {
    get "/source/data"
    store response -> raw { key: .id }
  }

  action Transform {
    for item in raw {
      map item -> CleanedItem { ... }
      store response -> cleaned { key: .id }
    }
  }

  action Load {
    for item in cleaned {
      post "/destination/data" { body: item }
    }
  }

  run Extract then Transform then Load
}

Each stage checkpoints independently. A failure in Load doesn't require re-running Extract or Transform.

Scheduled missions with interruption handling

mission ScheduledSync {
  checkpoint: afterStep
  schedule: cron("0 2 * * *")  // 2 AM daily

  action Sync {
    get "/updates" { since: lastSync }
    store response -> updates { key: .id }
  }

  run Sync
}

If the server restarts mid-sync, the mission resumes automatically.

Combining with other features

With trace

mission DebugableDurable {
  checkpoint: afterStep
  trace: full

  action Process {
    // Full state capture for both durability and debugging
  }

  run Process
}

With pause

mission LongRunning {
  checkpoint: afterStep

  action Step1 {
    get "/data"
    store response -> data { key: .id }
  }

  action WaitForApproval {
    pause {
      duration: "7d",
      resumeOn: webhook "/approved"
    }
  }

  action Step2 {
    for item in data {
      post "/process" { body: item }
    }
  }

  run Step1 then WaitForApproval then Step2
}

Programmatic access

import { execute, getExecutionState } from 'reqon';

// Check for existing checkpoint
const state = await getExecutionState('SyncMission');

if (state?.status === 'interrupted') {
  // Resume from checkpoint
  const result = await execute(source, {
    resume: true,
    executionId: state.executionId
  });
} else {
  // Fresh execution
  const result = await execute(source);
}

Best practices

1. Use afterStep for critical data - When data loss is unacceptable
2. Use onFailure for performance - When failures are rare and re-processing is cheap
3. Combine with retry - Checkpoints + retry provides comprehensive fault tolerance
4. Monitor checkpoint size - Large variable state increases checkpoint overhead

Performance considerations

Mode	Overhead	Durability
afterStep	Higher (checkpoint every step)	Maximum
onFailure	Lower (checkpoint only on failure)	Good
None	None	No durability

For high-throughput pipelines, consider:

mission HighThroughput {
  checkpoint: onFailure  // Lower overhead

  action Batch {
    for batch in batches checkpoint every 100 {
      // Checkpoint every 100 iterations, not every item
    }
  }

  run Batch
}