Troubleshooting Run Script for self-hosted

Troubleshooting 

This guide contains some common commands and areas to check in order to troubleshoot any setup issues with your self-hosted run script deployment using tines-command-runner and tines-app.

docker-compose 

exec into tines-command-runner container

docker ps (grab the <tcr_container_id>)
docker exec -it <tcr_container_id> /bin/bash

Review docker logs for tines-command-runner

docker logs -f <tcr_container_id>

Stop / Start all services

docker compose down && docker compose up -d

Stop only tines-command-runner service

docker stop <tcr_container_id>
docker start <tcr_container_id>

Review docker env variables for tines-command-runner:

docker inspect <tcr_container_id> | jq '.[0].Config.Env'

or if jq is not installed:

docker inspect <tcr_container_id> | grep -A 20 "Env"

Environment Variables 

tines-command-runner

  • PIP_INDEX_URL: Specifies the primary Python Package Index. Default is https://pypi.org/simple. This can be overridden to use a custom package index.

  • PIP_EXTRA_INDEX_URL: Allows specifying a fallback package index. Default is http://pypi-server:8080/simple/, which is the default local PyPI server(pypi-server) running alongside the command runner.

  • NO_PIP_INDEX: If set, disables all package indexes, including PIP_INDEX_URL and PIP_EXTRA_INDEX_URL, relying only on packages included in the container.

  • TRUSTED_HOST: Specifies a trusted package index host. Default is pypi-server, ensuring that the local PyPI server is trusted.

tines-app

  • TINES_COMMAND_RUNNER_PORT - defaults to 4400

  • TINES_COMMAND_RUNNER_HOST - defaults to tines-command-runner

capacity configuration

tines-command-runner can be configured to handle different levels of concurrent requests through two key environment variables:

  • TINES_COMMAND_RUNNER_WORKERS - Controls the number of Puma worker processes. (Default is 2).

  • TINES_COMMAND_RUNNER_THREADS - Controls the number of threads per worker. (Default is 5).

These values determine the total concurrent requests the command runner can handle (workers × threads). For example, with default values, the service can handle 10 concurrent requests (2 workers × 5 threads).

APP_USER 

Ensuring APP_USER can sudo into Users

As part of the permissions structure for run script we generate users within in the container via the Dockerfile.

This has a number of variables which can be overridden if necessary:

APP_USER=${APP_USER:-2000}
NUM_USERS=${NUM_USERS:-1000}
GROUP_ID=${GROUP_ID:-1000}
GROUP_NAME=${GROUP_NAME:-tines-team-group}

Check User Existence in /etc/passwd

getent passwd | grep tines-tcr-team-user

Check for Users Created with Specific UID Range:

awk -F: '$3 >= 2000 && $3 < 3000 {print $1, $3}' /etc/passwd

Verify Home Directories Exist

ls -ld /home/tines-tcr-team-user-*

Check Group Membership

getent group tines-team-group

Check file write permissions for a user (in the case of permissions errors)

sudo -u tines-tcr-team-user-2000 touch /specify-the-dir-path

UID Mappings 

Check permissions on team-uid-mappings

ls -ld /team-uid-mappings
cat /team-uid-mappings

View UID to Username Mapping (for All Users)

getent passwd

Check UID for a Specific User

id tines-tcr-team-user-2000

Connectivity 

This command should be run from within the tines-app container. It will attempt to connect to the tines-command-runner in a way that’s very similar to how the tines-app itself does it. It loads the host and port and then makes a POST request to the endpoint.

curl -vvv "http://${TINES_COMMAND_RUNNER_HOST:-tines-command-runner}:${TINES_COMMAND_RUNNER_PORT:-4400}/run_script" \
    -H "Content-Type: application/json" \
    -d '{
        "action_id": 1,
        "script_type": "python",
        "script_content": "def main(input_data):\n    print(f\"Received input_data: {input_data}\")\n    result = \"2\" + str(2)\n    return \"result is \" + str(result)\n",
        "pre_run_command": "uv pip install $TINES_DEPENDENCIES_LIST",
        "dependencies": [],
        "environment_id": "1234",
        "team_id": 100,
        "timeout": 60
    }'
Was this helpful?