Scroll to navigation

makeflow(1) Cooperative Computing Tools makeflow(1)


makeflow - workflow engine for executing distributed workflows


makeflow [options] <dagfile>


Makeflow is a workflow engine for distributed computing. It accepts a specification of a large amount of work to be performed, and runs it on remote machines in parallel where possible. In addition, Makeflow is fault-tolerant, so you can use it to coordinate very large tasks that may run for days or weeks in the face of failures. Makeflow is designed to be similar to Make, so if you can write a Makefile, then you can write a Makeflow.

You can run a Makeflow on your local machine to test it out. If you have a multi-core machine, then you can run multiple tasks simultaneously. If you have a Condor pool or a Sun Grid Engine batch system, then you can send your jobs there to run. If you don't already have a batch system, Makeflow comes with a system called Work Queue that will let you distribute the load across any collection of machines, large or small. Makeflow also supports execution in a Docker container, regardless of the batch system used.


When makeflow is ran without arguments, it will attempt to execute the workflow specified by the Makeflow dagfile using the local execution engine.


Clean up: remove logfile and all targets. If option is one of [intermediates, outputs, cache], only indicated files are removed.
Write summary of workflow to file.
 -h, --help
Show this help screen.
 -v, --version
Show version string.
Chdir to enable executing the Makefile in other directory.
 --argv <file>
Use command line arguments from JSON file.

Workflow Handling

 -a, --advertise
Advertise the manager information to a catalog server.
Use this file for the makeflow log. (default is X.makeflowlog)
Use this file for the batch system log. (default is X.<type>log)
Email summary of workflow to address.
Max number of local jobs to run at once. (default is # of cores)
Max number of remote jobs to run at once. (default is 1000 for -Twq, 100 otherwise)
 -R, --retry
Automatically retry failed batch jobs up to 100 times.
Automatically retry failed batch jobs up to n times.
 --local-cores <#>
Max number of cores used for local execution.
 --local-memory <#>
Max amount of memory used for local execution.
 --local-disk <#>
Max amount of disk used for local execution.


Batch Options

Add these options to all batch submit files.
Send all local environment variables in remote execution.
 --wait-for-files-upto <#>
Wait for output files to be created upto this many seconds (e.g., to deal with NFS semantics).
Time to retry failed batch job submission. (default is 3600s)
Batch system type: local, dryrun, condor, sge, pbs, torque, blue_waters, slurm, moab, cluster, wq, amazon, mesos. (default is local)
Excludes resources at submission. (SLURM, TORQUE, and PBS)
Excludes memory at submission. (SLURM)
 --batch-mem-type <type>
Specify memory resource type. (SGE)
 --working-dir <dir|url>
Working directory for batch system.
Run task in sandbox using bash script and task directory.
Set the job name based on the command.

JSON/JX Options

Interpret <dagfile> as a JSON format Makeflow.
Evaluate JX expressions in <dagfile>. Implies --json.
 --jx-args <args>
Read variable definitions from the JX file <args>.
 --jx-define <VAL=EXPR>
Set the variable <VAL> to the JX expression <EXPR>.
 --jx-context <ctx>
Deprecated. See '--jx-args'.

Debugging Options

Enable debugging for this subsystem.
Write debugging output to this file. By default, debugging is sent to stderr (":stderr"). You may specify logs to be sent to stdout (":stdout") instead.
 --debug-rotate-max <byte>
Rotate debug file once it reaches this size.
Display runtime progress on stdout.

WorkQueue Options

Set catalog server to <catalog>. Format: HOSTNAME:PORT
WorkQueue fast abort multiplier. (default is deactivated)
Set the project name to <project>.
Port number to use with WorkQueue. (default is 9123, 0=arbitrary)
Select port at random and write it to this file. (default is disabled)
Priority. Higher the value, higher the priority.
Work Queue keepalive timeout (default: 30s)
Work Queue keepalive interval (default: 120s)
WorkQueue scheduling algorithm. (time|files|fcfs)
 password <pwfile>
Password file for authenticating workers.
Disable file caching (currently only Work Queue, default is false)
 --work-queue-preferred-connection <connection>
Indicate preferred connection. Chose one of by_ip or by_hostname. (default is by_ip)

Monitor Options

 --monitor <dir>
Enable the resource monitor, and write the monitor logs to <dir>
 --monitor-exe <file>
Specify resource monitor executable.
Enable monitor time series. (default is disabled)
Enable monitoring of openened files. (default is disabled)
 --monitor-interval <#>
Set monitor interval to <#> seconds. (default 1 second)
 --monitor-log-fmt <fmt>
Format for monitor logs. (default resource-rule-%06.6d, %d -> rule number)
Monitor measures the task's current execution directory size.
 --allocation <waste>
When monitoring is enabled, automatically assign resource allocations to tasks. Makeflow will try to minimize waste or maximize throughput.

Umbrella Options

 --umbrella-binary <filepath>
Umbrella binary for running every rule in a makeflow.
 --umbrella-log-prefix <filepath>
Umbrella log file prefix for running every rule in a makeflow. (default is <makefilename>.umbrella.log)
 --umbrella-mode <mode>
Umbrella execution mode for running every rule in a makeflow. (default is local)
 --umbrella-spec <filepath>
Umbrella spec for running every rule in a makeflow.

Docker Support

 --docker <image>

Run each task in the Docker container with this name. The image will be obtained via "docker pull" if it is not already available.
 --docker-tar <tar>

Run each task in the Docker container given by this tar file. The image will be uploaded via "docker load" on each execution site.
 --docker-opt <string>
Specify options to be used in DSingularityocker execution.

Singularity Support

 --singularity <image>

Run each task in the Singularity container with this name. The container will be created from the passed in image.
 --singularity-opt <string>
Specify options to be used in Singularity execution.

Amazon Options

 --amazon-config <path>

Path to Amazon EC2 configuration file generated by makeflow_ec2_setup.

Amazon Lambda Options

 --lambda-config <path>

Path to Amazon Lambda configuration file generated by makeflow_lambda_setup.

Amazon Batch Options

 --amazon-batch-config <path>

Path to Amazon Batch configuration file generated by makeflow_amazon_batch_setup.
 --amazon-batch-img <img>

Specify image used for Amazon Batch execution.

Mesos Options

 --mesos-master <hostname>

Indicate the host name of preferred mesos manager.
 --mesos-path <filepath>

Indicate the path to mesos python2 site-packages.
 --mesos-preload <library>

Indicate the linking libraries for running mesos..

Kubernetes Options

 --k8s-image <docker_image>

Indicate the Docker image for running pods on Kubernetes cluster.

Mountfile Support

 --mounts <mountfile>
Use this file as a mountlist. Every line of a mountfile can be used to specify the source and target of each input dependency in the format of target source (Note there should be a space between target and source.).
 --cache <cache_dir>
Use this dir as the cache for file dependencies.

Archiving Options

 --archive <>
Archive results of workflow at the specified path (by default /tmp/makeflow.archive.$UID) and use outputs of any archived jobs instead of re-executing job
 --archive-dir <path>
Specify archive base directory.
 --archive-read <path>
Only check to see if jobs have been cached and use outputs if it has been
 --archive-s3 <s3_bucket>
Base S3 Bucket name
 --archive-s3-no-check <s3_bucket>
Blind upload files to S3 bucket (No existence check in bucket).
 --s3-hostname <s3_hostname>
Base S3 hostname. Used for AWS S3.
 --s3-keyid <s3_key_id>
Access Key for cloud server. Used for AWS S3.
 --s3-secretkey <secret_key>
Secret Key for cloud server. Used for AWS S3.

VC3 Builder Options

Enable VC3 Builder
 --vc3-exe <file>
VC3 Builder executable location
 --vc3-log <file>
VC3 Builder log name
 --vc3-options <string>
VC3 Builder option string

Other Options

 -A, --disable-afs-check
Disable the check for AFS. (experts only)
 -z, --zero-length-error
Force failure on zero-length output files.
Enable garbage collection. (ref_cnt|on_demand|all)
 --gc-size <int>
Set disk size to trigger GC. (on_demand only)
Set number of files to trigger GC. (ref_cnt only)
 --wrapper <script>

Wrap all commands with this script. Each rule's original recipe is appended to script or replaces the first occurrence of {} in script.
 --wrapper-input <file>

Wrapper command requires this input file. This option may be specified more than once, defining an array of inputs. Additionally, each job executing a recipe has a unique integer identifier that replaces occurrences %% in file.
 --wrapper-output <file>

Wrapper command requires this output file. This option may be specified more than once, defining an array of outputs. Additionally, each job executing a recipe has a unique integer identifier that replaces occurrences %% in file.
Use Parrot to restrict access to the given inputs/outputs.
 --parrot-path <path>
Path to parrot_run executable on the host system.
 --env-replace-path <path>
Path to env_replace executable on the host system.
Do not check for file existence before running.
Disable saving failed nodes to directory for later analysis.
 --shared-fs <dir>
Assume the given directory is a shared filesystem accessible at all execution sites.
Change to <dir> prior to executing the workflow.

MPI Options

 --mpi-cores <#>
Number of cores each MPI worker uses.
 --mpi-memory <#>
Amount of memory each MPI worker uses.
 --mpi-task-working-dir <path>
Path to the MPI tasks working directory.


When the batch system is set to -T <dryrun>, Makeflow runs as usual but does not actually execute jobs or modify the system. This is useful to check that wrappers and substitutions are applied as expected. In addition, Makeflow will write an equivalent shell script to the batch system log specified by -L <logfile>. This script will run the commands in serial that Makeflow would have run. This shell script format may be useful for archival purposes, since it does not depend on Makeflow.


When cctools is built with --with-mpi-path=which mpicc` configuration, Makeflow can be ran as an MPI program. To do so, run Makeflow as an argument to BOLD(mpirun)/BOLD(mpiexec) and set BOLD(-T) PARAM(mpi) as a Makeflow option. When submitting mpi, request one process per core. Makeflow will count up how many processes each node given to MPI has, and use that as the core count for the worker on that node. Makeflow will then share memory evenly amongst the cores on the node, following the following equation BOLD(worker_memory) = (BOLD(total_memory) / BOLD(total_logical_cores)) * BOLD(num_cores_for_worker). To override Makeflow sharing memory equally, or setting per-worker cores value, use OPTION_ITEM('--mpi-cores) and .TP  '--mpi-memory'

Tasks can also have their own sandbox. To specify the directory for tasks to create their sandbox subdirectory in, use .TP  '--mpi-task-working-dir'


The following environment variables will affect the execution of your Makeflow:


This corresponds to the -B <options> parameter and will pass extra batch options to the underlying execution engine.


This corresponds to the -j <#> parameter and will set the maximum number of local batch jobs. If a -j <#> parameter is specified, the minimum of the argument and the environment variable is used.


This corresponds to the -J <#> parameter and will set the maximum number of local batch jobs. If a -J <#> parameter is specified, the minimum of the argument and the environment variable is used.

Note that variables defined in your Makeflow are exported to the environment.


Inclusive low port in range used with -p 0.


Inclusive high port in range used with -p 0.


On success, returns zero. On failure, returns non-zero.


Run makeflow locally with debugging:

makeflow -d all Makeflow

Run makeflow on Condor will special requirements:

makeflow -T condor -B "requirements = MachineGroup == 'ccl'" Makeflow

Run makeflow with WorkQueue using named workers:

makeflow -T wq -a -N Makeflow

Create a directory containing all of the dependencies required to run the specified makeflow

makeflow -b bundle Makeflow


The Cooperative Computing Tools are Copyright (C) 2005-2019 The University of Notre Dame. This software is distributed under the GNU General Public License. See the file COPYING for details.