Users’ Manual

• HTCondor Quick Start Guide
  • A First HTCondor Job
  • The science Job Example
  • Expanding the science Job and the Organization of Files
  • Where to Go from Here
• Welcome and Introduction to HTCondor
• Running a Job: the Steps To Take
• Submitting a Job
  • Sample submit description files
  • Submitting many similar jobs with one queue command
  • Variables in the Submit Description File
  • Including Submit Commands Defined Elsewhere
  • Using Conditionals in the Submit Description File
  • Function Macros in the Submit Description File
  • About Requirements and Rank
  • Submitting Jobs Using a Shared File System
  • Jobs That Require Credentials
  • Jobs That Require GPUs
  • Interactive Jobs
  • Submitting Lots of Jobs
• Submitting Jobs Without a Shared File System: HTCondor’s File Transfer Mechanism
  • Specifying If and When to Transfer Files
  • Specifying What Files to Transfer
  • File Paths for File Transfer
  • Dataflow Jobs
  • Public Input Files
  • Behavior for Error Cases
  • File Transfer Using a URL
• Managing a Job
  • Checking on the progress of jobs
  • Peeking in on a running job’s output files
  • Starting an interactive shell next to a running job on a remote machine
  • Removing a job from the queue
  • Placing a job on hold
  • Changing the priority of jobs
  • Why is the job not running?
  • Job in the Hold State
  • In the Job Event Log File
  • Job Termination
  • Job Completion
  • Summary of all HTCondor users and their jobs
• Automatically managing a job
  • Automatically rerunning a failed job
  • Automatically removing a job in the queue
  • Automatically placing a job on hold
  • Automatically releasing a held job
  • Holding a completed job
• Services for Running Jobs
  • Environment Variables
  • Extra Environment Variables HTCondor sets for Jobs
  • Communicating with the Submit machine via Chirp
  • When changes to a job made by chirp take effect
  • Resource Limitations on a Running Job
• Priorities and Preemption
  • Job Priority
  • User priority
  • Details About How HTCondor Jobs Vacate Machines
• DAGMan Workflows
  • Describing Workflows with DAGMan
  • Node Job Submit File Contents
  • DAG Submission
  • File Paths in DAGs
  • DAG Monitoring
  • Editing a Running DAG
  • Removing a DAG
  • Suspending a Running DAG
  • Advanced Features of DAGMan
  • The Rescue DAG
  • DAG Recovery
  • Visualizing DAGs with dot
  • Capturing the Status of Nodes in a File
  • A Machine-Readable Event History, the jobstate.log File
  • Status Information for the DAG in a ClassAd
  • Managing Large Numbers of Jobs with DAGMan
  • Workflow Metrics
  • DAGMan and Accounting Groups
• Job Sets
  • Submitting a job set
  • Listing job sets
  • Checking on the progress of job sets
  • Removing a job set
• Matchmaking with ClassAds
  • Inspecting Machine ClassAds with condor_status
• Choosing an HTCondor Universe
  • Vanilla Universe
  • Grid Universe
  • Java Universe
  • Scheduler Universe
  • Local Universe
  • Parallel Universe
  • VM Universe
  • Docker Universe
  • Container Universe
• Java Applications
  • A Simple Example Java Application
  • Less Simple Java Specifications
  • Chirp I/O
• Parallel Applications (Including MPI Applications)
  • How Parallel Jobs Run
  • Parallel Jobs and the Dedicated Scheduler
  • Submission Examples
  • MPI Applications Within HTCondor’s Vanilla Universe
• Virtual Machine Applications
  • The Submit Description File
  • Checkpoints
  • Disk Images
  • Job Completion in the vm Universe
  • Failures to Launch
• Docker Universe Applications
  • Docker and Networking
• Container Universe Jobs
• Self-Checkpointing Applications
  • How To Run Self-Checkpointing Jobs
  • Requirements
  • Using checkpoint_exit_code
  • How Frequently to Checkpoint
  • Debugging Self-Checkpointing Jobs
  • Working Around the Assumptions
  • Other Options
  • Signals
• Time Scheduling for Job Execution
  • Job Deferral
  • CronTab Scheduling
• Special Environment Considerations
  • AFS
  • NFS
  • HTCondor Daemons That Do Not Run as root
  • Job Leases
  • Heterogeneous Submit: Execution on Differing Architectures
• Potential Problems
  • Renaming of argv[0]