we are a group of 3 developers and we have a shared server/driver that runs the prefect server (its ip is

10.10.10.255

) I am using the following docker-compose.yml (which I got and revised a bit by doing

prefect server config > docker-compose.yml

networks:
  prefect-server:
    name: prefect-server
services:
  apollo:
    command: bash -c "./post-start.sh && npm run serve"
    depends_on:
    - graphql
    - hasura
    environment:
      GRAPHQL_SERVICE_HOST: <http://graphql>
      GRAPHQL_SERVICE_PORT: 4201
      HASURA_API_URL: <http://hasura:3000/v1alpha1/graphql>
      PREFECT_API_HEALTH_URL: <http://graphql:4201/health>
      PREFECT_API_URL: <http://graphql:4201/graphql/>
      PREFECT_SERVER__TELEMETRY__ENABLED: "true"
    healthcheck:
      interval: 10s
      retries: 60
      start_period: 1s
      test: curl --fail --silent "<http://apollo:4200/.well-known/apollo/server-health>"
        &> /dev/null || exit 1
      timeout: 2s
    image: prefecthq/apollo:core-1.2.2
    networks:
      prefect-server: null
    ports:
    - 0.0.0.0:4200:4200/tcp
    restart: always
  graphql:
    command: bash -c "prefect-server database upgrade -y && python src/prefect_server/services/graphql/server.py"
    depends_on:
    - hasura
    environment:
      PREFECT_CORE_VERSION: 1.2.2
      PREFECT_SERVER_DB_CMD: prefect-server database upgrade -y
      PREFECT_SERVER__DATABASE__CONNECTION_URL: <postgresql://prefect>:test-password@postgres:5432/prefect_server
      PREFECT_SERVER__HASURA__ADMIN_SECRET: hasura-secret-admin-secret
      PREFECT_SERVER__HASURA__HOST: hasura
    healthcheck:
      interval: 20s
      retries: 60
      start_period: 1s
      test: curl --fail --silent "<http://graphql:4201/health>" &> /dev/null || exit
        1
      timeout: 2s
    image: prefecthq/server:core-1.2.2
    networks:
      prefect-server: null
    ports:
    - 0.0.0.0:4201:4201/tcp
    restart: always
  hasura:
    command: graphql-engine serve
    depends_on:
    - postgres
    environment:
      HASURA_GRAPHQL_DATABASE_URL: <postgresql://prefect>:test-password@postgres:5432/prefect_server
      HASURA_GRAPHQL_ENABLE_CONSOLE: "true"
      HASURA_GRAPHQL_LOG_LEVEL: warn
      HASURA_GRAPHQL_QUERY_PLAN_CACHE_SIZE: 100
      HASURA_GRAPHQL_SERVER_PORT: '3000'
      HASURA_GRAPHQL_V1_BOOLEAN_NULL_COLLAPSE: "true"
    healthcheck:
      interval: 10s
      retries: 60
      start_period: 1s
      test: wget -O - <http://hasura>:$${HASURA_GRAPHQL_SERVER_PORT}/healthz &>/dev/null
        || exit 1
      timeout: 2s
    image: hasura/graphql-engine:v2.1.1
    init: true
    networks:
      prefect-server: null
    ports:
    - 0.0.0.0:3000:3000/tcp
    restart: always
  postgres:
    command:
    - postgres
    - -c
    - max_connections=150
    environment:
      POSTGRES_DB: prefect_server
      POSTGRES_PASSWORD: test-password
      POSTGRES_USER: prefect
    healthcheck:
      interval: 10s
      retries: 60
      start_period: 2s
      test: pg_isready -q -d $${POSTGRES_DB} -U $${POSTGRES_USER} || exit 1
      timeout: 2s
    image: postgres:11
    networks:
      prefect-server: null
    ports:
    - 0.0.0.0:5432:5432/tcp
    restart: always
  towel:
    command: python src/prefect_server/services/towel/__main__.py
    depends_on:
    - graphql
    environment:
      PREFECT_SERVER__HASURA__ADMIN_SECRET: hasura-secret-admin-secret
      PREFECT_SERVER__HASURA__HOST: hasura
    image: prefecthq/server:core-1.2.2
    networks:
      prefect-server: null
    restart: always
  ui:
    command: /intercept.sh
    depends_on:
    - apollo
    environment:
      PREFECT_SERVER__APOLLO_URL: <http://10.10.10.255:4200/graphql>
    healthcheck:
      interval: 30s
      retries: 3
      test: curl --fail --silent --head "<http://ui:8080/>" &> /dev/null || exit 1
      timeout: 5s
    image: prefecthq/ui:core-1.2.2
    networks:
      prefect-server: null
    ports:
    - 0.0.0.0:8080:8080/tcp
    restart: always
version: '3.7'

The issue is that the UI when accessed from developer browser can't access the graphql endpoint. How can I tell the server to look for the API elsewhere? Currently it tries to access it on http://localhost:4200/graphql by default I saw on this github thread some mentioning of

VUE_APP_SERVER_URL

but I am not certain how to fit it in.

You need to set it

[server]
  
  [server.ui]
    
    apollo_url = "<http://YOUR_MACHINES_PUBLIC_IP:4200/graphql>"

before you start the server in the config.toml or under UI to point to the API

@Kevin Kho thanks, where are those .toml files located?

you can create the config.toml if it doesnt exist

not sure where though as I have the docker-compose.yml and I can't see any volume or any external dirs it depends on

Ah yeah you are right it’s the UI URL seeming to not stick. Are you making other changes that you need to start from

docker-compose

yourself rather than using

prefect server start

?