Skip to content

Latest commit

 

History

History
178 lines (104 loc) · 5.68 KB

README.rst

File metadata and controls

178 lines (104 loc) · 5.68 KB
Raw

HTCondor File Transfer Execution Point

HTCondor execution point setup and configuration for using HTCondor file transfer to synchronize a directory between two hosts.

Intended for use with xfer.py.

Requirements

  • The ability to run software as a user that can read and write files in the directory being synchronized.
  • The hostname of the central manager to which the execution point should report and the ability to contact the central manager's administrators.
  • Outgoing network connectivity to the central manager.
  • The name of the owner of the transfer jobs that should run on the EP.

Quickstart: Running the EP as a container

In this setup, the EP runs as a container using Docker or Singularity. Configuration and the directory to synchronize are provided via environment variables and volume mounts.

  1. Clone this repository:

    git clone https://github.com/brianaydemir/htcondor_file_transfer_ep.git

  2. Request a token, which the EP will use for authentication:

    cd htcondor_file_transfer_ep

./request_token.sh -c <central manager's hostname>

    For the CHTC, note that the central manager's hostname is "cm.chtc.wisc.edu".

    Run the script without any arguments to see all of the available options.

    Note that this command will hang until the token request is approved. Run it while in contact with the central manager's administrators.

    Take note of the request ID, and if you are CHTC staff with root on the CM, you can approve the token request yourself: log into cm.chtc.wisc.edu and as root/sudo run:

    condor_token_request_approve -reqid ID

    If you lost the request ID, you can see what's pending there with:

    condor_token_request_list

  3. Copy .env.template to .env and update it for your setup.

    You will need to edit your .env file. At a minimum, you need to set:

    CENTRAL_MANAGER=cm.chtc.wisc.edu
UNIQUE_NAME=pick-a-unique-name-here
JOB_OWNER=your-user-on-the-AP

    Also consider setting DATA_DIR to a directory on the HOST where you will run the EP, which will be mounted onto /data inside the EP container.

  4. Start the EP as a Docker container:

    ./run_docker.sh [optional name for the container]

    Or as a Singularity container:

    ./run_singularity.sh [optional name for the instance]

  5. Use docker container stop or singularity instance stop to stop the container.

Quickstart: Running the EP directly on a host

In this setup, the EP runs directly on a host. Configuration is specified via the setup script (see below), and the directory to synchronize must be directly accessible via the host's file system.

  1. Clone this repository:

    git clone https://github.com/brianaydemir/htcondor_file_transfer_ep.git

  2. Start the setup script:

    cd htcondor_file_transfer_ep

./setup.sh -c <central manager's hostname> -n <unique name for this EP> -u <name of the transfer jobs' owner>

    Run the script without any arguments to see all of the available options.

  3. When prompted by the script, ask an administrator of the central manager to approve the given token request ID.

  4. After the script completes successfully, start the execution point by following the instructions in the script's final output.

Quickstart: Transferring files between EP and AP at CHTC

  1. Take a minute to note the distro version inside the Dockerfile; it should have a line like:

    FROM htcondor/mini:9.10-el7

    The "el7" there corresponds to "OpSysMajorVer == 7". You may need to add this later to your transfer job requirements in order for your transfer job to match properly.

  2. Log into your AP (eg submit2.chtc.wisc.edu).

    You should be able to find a slot for your running container like so:

    condor_status | grep your-unique-name-here

    The slot name should look something like:

    "slot1@your-unique-name-here@conainerhash"

    If that's all good you can now submit the transfer job...

  3. The tool repo to clone is:

    https://github.com/HTPhenotyping/htcondor_file_transfer

    Clone this repo locally and enter the repo checkout.

  4. Create subdirs for logs, and source and/or dest.

    Make a "working dir" locally for all the condor logs to live; eg:

    mkdir working

    Make a destination dir, if you are pulling files from the EP; eg:

    mkdir dest

    Make a source dir, if you are pushing files to the EP; eg:

    mkdir src   # and put stuff in it

    If you want to pull files from the DATA_DIR on your host, remember that this is mounted to /data inside the EP container.

  5. Submit the transfer job.

    The command line to start the transfer job to sync your DATA_DIR from your host to your newly-created dest dir on the AP would be:

    req='UniqueName == "mystras-xfer-ep" && OpSysMajorVer == 7'
./xfer.py sync --requirements "$req" \
               --working-dir working/ pull dest/ /data

    And again note that OpSysMajorVer should match what you found in the Dockerfile from the EP container setup (in htcondor_file_transfer_ep).

    To go the other way, to send a src dir on the AP to the DATA_DIR on you host, use the "push" command instead:

    req='UniqueName == "mystras-xfer-ep" && OpSysMajorVer == 7'
./xfer.py sync --requirements "$req" \
               --working-dir working/ push src/ /data

    You can watch progress on the AP by doing condor_q, or watching the logs under your new "working" dir. It may take some time for everything to finish.