README.md

Retry mechanism

To enable Nextflow's process retry mechanism, set up the configuration for process-specific resource allocation by adding a retry_strategy namespace for each process that requires a retry mechanism. In these namespaces, a retry strategy for memory and cpus can be specified (both are not required, include only the resource(s) that require an update with each retry) as separate namespaces, where each must define a strategy (the mathematical operation for updating the resource) and an operand (the value used to perform the strategy).

Example:

process {
    withName: 'proc1' {
        cpus = 4
        memory = 8.GB
        retry_strategy {
            memory {
                strategy = 'exponential'
                operand = 2
            }
            cpus {
                strategy = 'add'
                operand = 6
            }
        }
    }
}

Available strategies:

• add - add operand with each retry
  • With this strategy, the operand must match the units of the corresponding resource. Ex. for memory, the operand must be a MemoryUnit (ex. 1.GB)
• subtract - subtract operand with each retry
  • With this strategy, the operand must match the units of the corresponding resource. Ex. for memory, the operand must be a MemoryUnit (ex. 1.GB)
• exponential - multiply by operand with each retry, results in exponential growth with operand > 1 and exponential decay with 0 < operand < 1.

Example

To convert these configurations into Nextflow retry settings, call the setup_retry function from the retry namespace in methods.config.

Example:

includeConfig "/path/to/retry.config"
...
methods {
    ...
    setup = {
        ...
        retry.setup_retry()
    }
}