Migrating from Deprecated Attributes

This guide explains how to migrate from the deprecated skip and retryable_errors attributes to their modern, more powerful replacements.

Terragrunt has deprecated two attributes in favor of more flexible block-based configurations:

• skip → Use the exclude block instead
• retryable_errors → Use the errors block with retry sub-blocks instead

These new blocks provide more granular control and composability compared to the simple attributes they replace.

Migrating from skip to exclude

Why was skip deprecated?

The skip attribute was a simple boolean that would exclude a unit from the run queue. The new exclude block provides much more flexibility:

• Exclude the unit only for specific OpenTofu/Terraform commands (e.g., only plan but not apply)
• Use conditional logic to determine when to exclude the unit
• Combine multiple conditions
• Better integration with other Terragrunt features

Basic Migration

Before:

skip = true

After:

exclude {
  if = true
  actions = ["all"]
}

Conditional Skip

Before:

skip = get_env("ENVIRONMENT") == "production"

After:

exclude {
  if = get_env("ENVIRONMENT") == "production"
  actions = ["all"]
}

Skip Specific Actions

The new exclude block allows you to exclude the unit only for specific OpenTofu/Terraform commands:

exclude {
  if = get_env("SKIP_DESTROY") == "true"
  actions = ["destroy"]
}

This wasn't possible with the old skip attribute!

Migrating from retryable_errors to errors Block

Why was retryable_errors deprecated?

The retryable_errors attribute was a simple list of error patterns. The new errors block with retry sub-blocks provides:

• Multiple retry configurations with different patterns and settings
• Named retry blocks for better documentation
• Per-retry configuration of max attempts and sleep intervals
• Composability - combine multiple retry strategies
• Better organization for complex retry logic

Basic Migration

Before:

retryable_errors = [
  ".*Error: transient network issue.*",
  ".*Error: timeout.*"
]

retry_max_attempts     = 3
retry_sleep_interval_sec = 5

After:

errors {
  retry "transient_errors" {
    retryable_errors = [
      ".*Error: transient network issue.*",
      ".*Error: timeout.*"
    ]
    max_attempts = 3
    sleep_interval_sec = 5
  }
}

Using Default Retryable Errors

If you were using the get_default_retryable_errors() function:

Before:

retryable_errors = concat(
  get_default_retryable_errors(),
  [".*custom error.*"]
)

After:

errors {
  retry "default_errors" {
    retryable_errors = get_default_retryable_errors()
    max_attempts = 3
    sleep_interval_sec = 5
  }

  retry "custom_errors" {
    retryable_errors = [".*custom error.*"]
    max_attempts = 5
    sleep_interval_sec = 10
  }
}

Note: The get_default_retryable_errors() function still works and returns the default list for use within the errors block.

Multiple Retry Strategies

The new errors block allows you to define different retry strategies for different types of errors:

errors {
  # Quick retries for transient network issues
  retry "network_errors" {
    retryable_errors = [
      ".*connection reset.*",
      ".*timeout.*"
    ]
    max_attempts = 5
    sleep_interval_sec = 2
  }

  # Slower retries for rate limiting
  retry "rate_limit_errors" {
    retryable_errors = [
      ".*rate limit exceeded.*",
      ".*too many requests.*"
    ]
    max_attempts = 10
    sleep_interval_sec = 30
  }

  # Few retries for potential transient API issues
  retry "api_errors" {
    retryable_errors = [
      ".*internal server error.*"
    ]
    max_attempts = 3
    sleep_interval_sec = 15
  }
}

This level of granularity wasn't possible with the old retryable_errors attribute!

Error Messages

If you try to use the deprecated attributes, Terragrunt will fail with an HCL parsing error:

For skip attribute:

Error: Unsupported argument

  on terragrunt.hcl line 2:
   2: skip = true

An argument named "skip" is not expected here.

For retryable_errors attribute:

Error: Unsupported argument

  on terragrunt.hcl line 4:
   4: retryable_errors = [".*Error: transient.*"]

An argument named "retryable_errors" is not expected here.

These errors indicate that the attributes have been completely removed from Terragrunt. Please refer to the migration examples below.

How Retry Errors Are Collected

When you define multiple retry blocks within the errors block, Terragrunt automatically collects all the retryable_errors patterns from all retry blocks and uses them for error matching.

Example:

errors {
  retry "network_errors" {
    retryable_errors = [".*timeout.*", ".*connection reset.*"]
    max_attempts = 5
    sleep_interval_sec = 2
  }

  retry "api_errors" {
    retryable_errors = [".*rate limit.*", ".*429.*"]
    max_attempts = 10
    sleep_interval_sec = 30
  }
}

In this example, Terragrunt will retry on any error matching:

• .*timeout.*
• .*connection reset.*
• .*rate limit.*
• .*429.*

Each retry block can have its own max_attempts and sleep_interval_sec, allowing fine-grained control over retry behavior for different error types—for example, one block can retry at 2-second intervals while another uses 30-second intervals.

Further Reading

• Exclude Block Reference
• Errors Block Reference
• All Attributes Reference