Skip to content

Latest commit

 

History

History
92 lines (75 loc) · 4.39 KB

configuration_index.md

File metadata and controls

92 lines (75 loc) · 4.39 KB
Raw

Configuration Index

List of available configuration options.

[Mode]

  • debug: True/False. Set whether in debug mode or not (Don't know what debug does though...?)
  • dryrun: True/False. Doesn't actually submit jobs.

[Glidein]

  • address: Address of the glidein server, which keeps track of the global condor queue.
  • uuid: The UUID of the glidein.
  • site: Name of the site this config is for (ex. Cedar)
  • delay: How long to wait before considering the next set of jobs; if client.py should be run by cron, then use -1

[Cluster]

  • user: User under which jobs are being submitted.
  • os: The name of the OS on the cluster (ex. RHEL6, Ubuntu_16.04).
  • cvmfs: True/False. Whether the site has access to cvmfs or not.
  • limit_per_submit: The max number of jobs that can be submitted at once.
  • submit_command: Command used to submit jobs on this cluster (ex. condor_submit, sbatch...)
  • mem_only: True/False. Set only mem requirement for PBS (default: False).
  • vmem_only: True/False. Virtual memory only; applies to PBS-like clusters.
  • set_gpu_req: True/False. Set gpus requirement for PBS (default: True).
  • scheduler: Name of scheduler (ex. slurm, HTCondor, LSF...)
  • walltime_hrs: Max number of hours to run as integer (may not be enforced)
  • group_jobs: True/False. Submits multiple glideins with the same config.
  • running_cmd: Command needed to determine number of jobs running (ex. squeue ...).
  • partitions: ExamplePartition (Where ExamplePartition correspond to labelled section below). User-defined configurations included in this file.
  • pmem_only: True/False. Physical memory; a compliment to vmem_only.
  • pvmem: True/False. ???
  • prioritize_jobs: Array of strings representing priorities in order of priority (ex. ["memory", "disk"])
  • node_property: ???
  • cleanup: True/False. Whether to cleanup after finishing or not.
  • dir_cleanup: Absolute path to directory to clean.

[ExamplePartition]

  • cpu_only: True/False. If only CPUs are used or not.
  • gpu_only: True/False. If only GPUs are used or not.
  • max_cpus_per_job: Max number of CPUs to use for each job.
  • max_gpus_per_job: Max number of GPUs to use for each job.
  • max_memory_per_job: Max number of Megabytes of memory to use for each job.
  • max_disk_per_job: Max number of Megabytes of disk space to use per job.
  • whole_node: True/False. Use a fixed size of resources.
  • whole_node_cpus: The amount of CPUs for the glidein.
  • whole_node_gpus: The amount of GPUs for the glidein.
  • whole_node_memory: Number of Megabytes of memory for glidein.
  • mem_per_core: Number of Megabytes of memory per allocated per core.
  • whole_node_disk: Number of Megabytes of disk space.
  • running_cmd: Command needed to determine number of jobs running.
  • idle_cmd: Command needed to determine number of jobs idle.
  • max_total_jobs: Number of total jobs in any state as integer.
  • max_idle_jobs: Number of jobs that can be in idle state as integer.
  • limit_per_submit: Number of jobs that can be submitted per run as integer.
  • submit_command: Command to issue on submission.

[SubmitFile]

  • log: Absolute path to log file for submission.
  • filename: The name of the submit file (ex. submit.pbs)
  • env_wrapper_name: Name of the env_wrapper file to clean the environment (and sets variables).
  • requirements: List of requirements for submission (ex. CUDACapability)
  • executable: Absolute path to executable to be run.
  • custom_middle: Custom extra line(s) needed for cluster.
  • custom_end: Custom extra line(s) needed at the end of job.
  • mem_scale: Conversion from KB to MB, because LSF scheduler uses KB values but we need MB by default.
  • local_dir: Absolute path to temporary directory that will be used.
  • gpu_submit: GPU request command.
  • custom_header: To add something to the submit file BEFORE the automated entry (prefix on the submit file)
  • ref_host: ???
  • mem_safety_scale: ???

[StartdChecks]

  • enable_startd_checks: True/False. Enable or disable Startd monitoring checks.

[StartdLogging]

  • bucket: String. Name of S3 bucket
  • send_startd_logs: True/False. Send Startd Logs to S3 endpoint
  • url: String. URL of S3 endpoint.

[CustomEnv]

  • List custom environment variables here
  • http_proxy: The address to a required proxy.
  • SINGULARITY_BIN: The absolute path to singularity (for use in container configs)
  • CONTAINER: The absolute path to the singularity image to use.
  • GROUP: This corresponds to the Affiliation classad in the Condor job. This is for ranking ONLY. (ex. Alberta)