Continue On Conditions

The continue_on configuration allows workflows to continue execution even when steps encounter failures, specific exit codes, or produce certain outputs. This powerful feature enables resilient workflows that can handle errors gracefully and implement sophisticated control flow patterns.

Overview

By default, Dagu stops workflow execution when a step fails (returns a non-zero exit code). The continue_on configuration overrides this behavior, allowing you to:

• Continue execution after failures
• Handle specific exit codes differently
• React to command output patterns
• Mark steps as successful despite failures
• Build fault-tolerant workflows

Syntax

The continue_on field supports two syntaxes:

Shorthand Syntax

For simple cases, use a string value:

steps:
  - run: rm -rf /tmp/cache/*
    continue_on: failed    # Continue if step fails

  - run: echo "Optional"
    continue_on: skipped   # Continue if step is skipped

Object Syntax

For advanced configuration with multiple options:

steps:
  - run: echo "Complex case"
    continue_on:
      failure: true
      exit_code: [0, 1, 2]
      output: ["WARNING", "re:^INFO:.*"]
      mark_success: true

Configuration Fields

The continue_on configuration supports the following fields:

Field	Type	Description	Default
failure	boolean	Continue execution when the step fails	false
skipped	boolean	Continue execution when the step is skipped	false
exit_code	array	Continue execution for specific exit codes	[]
output	array	Continue execution when output matches patterns	[]
mark_success	boolean	Mark the step as successful when conditions are met	false

Field Details

failure

When set to true, the workflow continues even if the step fails with any non-zero exit code.

steps:
  # Shorthand syntax
  - id: optional_cleanup_shorthand
    run: rm -rf /tmp/cache/*
    continue_on: failed

  # Object syntax (equivalent)
  - id: optional_cleanup_object
    run: rm -rf /tmp/cache/*
    continue_on:
      failure: true

skipped

When set to true, the workflow continues when a step is skipped due to unmet preconditions.

steps:
  # Shorthand syntax
  - id: conditional_task_shorthand
    run: echo "Processing"
    preconditions:
      - condition: "${env.ENABLE_FEATURE}"
        expected: "true"
    continue_on: skipped

  # Object syntax (equivalent)
  - id: conditional_task_object
    run: echo "Processing"
    preconditions:
      - condition: "${env.ENABLE_FEATURE}"
        expected: "true"
    continue_on:
      skipped: true

exit_code

An array of specific exit codes that should not stop the workflow. This is useful when dealing with commands that use non-zero exit codes for non-error conditions.

steps:
  - id: check_service
    run: echo "Health check OK"
    continue_on:
      exit_code: [0, 1, 2]  # 0=healthy, 1=warning, 2=maintenance

output

An array of patterns to match against the command's stdout output. Supports both literal strings and regular expressions (prefixed with re:).

steps:
  - id: validate_data
    run: echo "Validating"
    continue_on:
      output:
        - "WARNING"                    # Literal string match (substring)
        - "SKIP"                       # Another literal string
        - "re:^INFO:.*"                # Regex: lines starting with "INFO:"
        - "re:WARN-[0-9]+"             # Regex: WARN- followed by numbers

Pattern Matching Rules:

• Literal patterns: Matched as substrings (e.g., "WARNING" matches "WARNING: Low memory")
• Regex patterns: Must start with re: prefix (e.g., re:^ERROR.*). A pattern that is not a valid regular expression is rejected when the workflow loads
• Patterns are matched against each line of stdout only (stderr is not checked)
• Matching is case-sensitive
• The num: numeric-comparison prefix is not supported here and is rejected when the workflow loads. Matching here reads one line at a time, while a numeric comparison reads a whole value. Numeric comparison applies to preconditions and router routes. See Numeric Comparison. To match output that begins with num:, write it as a regex, such as re:num:>=5

mark_success

When set to true, the step is marked as successful if any of the continue conditions are met, even if it would normally be considered a failure.

steps:
  - id: best_effort_optimization
    run: echo "Optimizing"
    continue_on:
      failure: true
      mark_success: true  # Step shows as successful in UI/logs

Common Patterns

Optional Steps

For steps that are nice-to-have but not critical:

steps:
  - id: cache_warmup
    run: echo "Warming cache"
    continue_on: failed

  - id: main_process
    run: echo "Processing"
    depends: cache_warmup

Handling Known Exit Codes

When working with tools that use exit codes for non-error states:

steps:
  - id: git_diff
    run: git diff --exit-code
    continue_on:
      exit_code: [0, 1]  # 0=no changes, 1=changes exist

  - id: process_changes
    run: echo "Handling changes"
    depends: git_diff

Warning Detection

Continue execution but handle warnings differently:

steps:
  - id: lint_code
    run: eslint .
    continue_on:
      output: ["WARNING", "re:.*warning.*"]
      exit_code: [0, 1]  # 0=no issues, 1=warnings only

  - id: strict_lint
    run: eslint . --max-warnings 0
    continue_on:
      failure: false  # This one must pass
    depends: lint_code

Graceful Degradation

Build workflows that degrade gracefully:

steps:
  - id: try_optimal_method
    run: echo "Processing with optimal settings"
    continue_on: failed

  - id: fallback_method
    run: echo "Processing with fallback settings"
    preconditions:
      - condition: "${try_optimal_method.exit_code}"
        expected: "re:[1-9][0-9]*"  # Only run if previous failed
    depends: try_optimal_method

Complex Output Matching

React to specific output patterns:

steps:
  - id: deployment_check
    run: kubectl rollout status deployment/app
    continue_on:
      output:
        - "re:Waiting for.*replicas"
        - "re:deployment.*not found"
        - "Unable to connect"
      exit_code: [1]

  - id: handle_deployment_issue
    run: echo "Fixing deployment"
    depends: deployment_check

Interaction with Other Features

With Retry Policies

continue_on is evaluated after all retries are exhausted:

steps:
  - id: flaky_service
    run: curl -fsS https://service.example.com/api
    retry_policy:
      limit: 3
      interval_sec: 5
    continue_on:
      exit_code: [22]  # Continue if curl still gets an HTTP error after retries

With Lifecycle Handlers

When a step with continue_on fails but the DAG continues, the final status is partially_succeeded, which triggers the onSuccess handler (not onFailure):

handler_on:
  success:
    run: 'echo "DAG completed (status: ${context.run.status})"'  # partially_succeeded

steps:
  - id: optional_step
    run: exit 1
    continue_on: failed  # Continues, DAG ends as partially_succeeded

With Dependencies

Dependent steps see the actual status unless mark_success is used:

steps:
  - id: step_a
    run: exit 1
    continue_on:
      failure: true
      mark_success: false  # Default

  - id: step_b
    run: 'echo "Step A status: failed"'
    depends: step_a  # Runs because of continue_on

  - id: step_c
    run: exit 1
    continue_on:
      failure: true
      mark_success: true  # Override status

  - id: step_d
    run: 'echo "Step C status: success"'
    depends: step_c  # Sees step-c as successful

Examples

Database Migration with Warnings

steps:
  - id: run_migration
    run: echo "Running migration"
    continue_on:
      output:
        - "re:WARNING:.*already exists"
        - "re:NOTICE:.*will be created"
      exit_code: [0, 1]  # 1 might indicate warnings

  - id: verify_migration
    run: echo "Verifying database"
    depends: run_migration

Multi-Cloud Deployment

steps:
  - id: deploy_aws
    run: echo "Deploying to AWS"
    continue_on: failed  # Continue even if AWS fails

  - id: deploy_gcp
    run: echo "Deploying to GCP"
    continue_on: failed  # Continue even if GCP fails

  - id: verify_deployment
    run: echo "Verifying cloud deployment"
    # No continue_on - at least one cloud must be working
    depends: [deploy_aws, deploy_gcp]

Service Health Check

steps:
  - id: check_primary
    run: curl -f https://primary.example.com/health
    continue_on:
      exit_code: [0, 22, 7]  # 22=HTTP error, 7=connection failed

  - id: check_secondary
    run: curl -f https://secondary.example.com/health
    preconditions:
      - condition: "${check_primary.exit_code}"
        expected: "re:[1-9][0-9]*"  # Only if primary failed
    depends: check_primary