https://prefect.io
Join Slack
Please help a person in despair Hello, I'm trying to setup Prefect inside a `docker-compose` that r...
# ask-community
v
Please help a person in despair Hello, I'm trying to setup Prefect inside a 
docker-compose
that runs on a remote VM. However, I cannot configure the correct URL for the 
graphql
(when opening the UI, I get the error 
Couldn't connect to Prefect Server at <http://localhost:4200/graphql>
) I tried to configure it from the 
config.toml
file with the URL of my machine:
Copy code
[server]
  [server.ui]
    graphql_url = "http://<<MACHINE_PUBLIC_IP>>:4200/graphql"
but it still tries the access localhost. Any help will be thoroughly appreciated. It's been 12 hours..
✅ 1
1️⃣ 1
r
Hi Vlad, what does your Docker Compose service for Prefect look like in your 
docker-compose.yml
file?
👀 1
j
Sorry to hear that. Please don’t despair. 🙂 Is this guide helpful?
❤️ 1
v
@Jeff Hale I've been using that already 😄
This is the 
docker-compose
service:
Copy code
ui:
    command: /intercept.sh
    depends_on:
      apollo:
        condition: service_healthy
    volumes:
      - type: bind
        source: ./config.toml
        target: /root/.prefect/config.toml
        read_only: true
    healthcheck:
      interval: 30s
      retries: 3
      test: curl --fail --silent --head "<http://ui:4001/>" &> /dev/null || exit 1
      timeout: 5s
    image: prefecthq/ui:core-1.2.2
    ports:
      - 4001:8080
    restart: always
    networks:
      - prefect-server
Now I question wether it's the correct wai to pass the 
config.toml
I'd prefer using environment variables but resources online are conflicting on the names of the variables
I now tried setting 
PREFECT__SERVER__UI__GRAPHQL_URL
(but also 
PREFECT_SERVER__UI__GRAPHQL_URL
) as env variables but still the UI check localhost
r
does the 
apollo
service in the Docker Compose yaml still have:
Copy code
ports:
      - 4200:4200/tcp
v
@Ryan PedenYes, moreover I can check that the port configuration is not correctly passed by looking at the calls the UI makes, and they are still on localhost
r
To make sure I'm understanding correctly - you want the UI to contact the server's IP instead of 
localhost
?
v
Yes, and I am not sure how to configure a non-default URL for graphql
In some resources I find:
Copy code
[server]
  [server.ui]
    apollo_url = "<http://YOUR_MACHINES_PUBLIC_IP:4200/graphql>"
In others:
Copy code
[server]
  [server.ui]
    graphql_url = "<http://YOUR_MACHINES_PUBLIC_IP:4200/graphql>"
r
In this pull request, it looks like an option was added to set a 
PREFECT_SERVER__APOLLO_URL
environment variable. You may need to run 
localStorage.clear()
in your browser's console afterward because it might have the old URL cached
v
Seems like 
PREFECT_SERVER__APOLLO_URL
woks but indeed, it's crucial to clear the cache of the browser. Gee that was a headache 😄 Thank you so much kind sir
Now I have simmilar issues when trying to connect the client to the Server 🙃
Copy code
PREFECT__SERVER__HOST: <http://apollo>
      PREFECT__SERVER__PORT: 4200
In the environment variables I should also replace apollo with my machine IP?
r
yes, I believe so - 
<http://apollo>
will only work if the client container is connected to the same Docker network as the server container
v
In this case, it is in the same network (and still doesn't seem to work as no agents and projects appear in the ui) If I configure
Copy code
PREFECT__SERVER__HOST: http://<<MACHINE_IP>>
      PREFECT__SERVER__PORT: 4200
I get 
ConnectionTimedOut
🥲
r
If you are using the client example from the repo Jeff linked above, did you change 
host
in 
config.toml
? When I set the host there, it works whether I use 
apollo
or the server IP as the host
v
not really, it keeps loading 😕 In the logs I see the 
<http://44.205.14.8:4001/>
error
Copy code
urllib3.exceptions.MaxRetryError: HTTPConnectionPool(host='3.205.14.8', port=4200): Max retries exceeded with url: / (Caused by ConnectTimeoutError(<urllib3.connection.HTTPConnection object at 0x7f98064740d0>
r
Is 
3.205.14.8
the IP of the server where you used docker compose to run the server?
v
yes
It does not fail if I pass 
<http://172.17.0.1>
but neither does it work
I mean, in the logs I see that the Gaent is created but I guess it is rregistred to a different database as it does not appear in the UI
@Ryan Peden Do I need to set up any other URL? for example 
PREFECT__API__URL=<http://postgres:5432>
?
In the logs I see that the flow is registred as 
<http://localhost:8080/default/flow/5957343a-9867-430d-b2f9-053cf764d57d>
, so this is not good, right?
r
I think that is okay. I see the same thing but the flow still shows up in my UI as expected
v
do you run on your local device? I did not have these issues also while running locally
P.S I also run things through an nginx, do you think it can cause the issue?
r
I believe so. It depends on how nginx is configured, which ports it is set up to receive connections on and forward to other services, etc. That is more a network/nginx config issue though, so it's beyond the scope of what I diagnose/troubleshoot
v
Either way, tahnk you for your patience. If I find the solution I'll post it here
r
Thanks, Vlad. I appreciate that. And I'm happy to help. Anything you find would be helpful. Eventually, I would like to create a tutorial on running Prefect using Docker Compose a Nginx...and although I focus mostly on Prefect 2 and not Prefect 1, I suspect that anything Nginx + docker-compose related will be applicable to both.
v
I think I got over some problems with nginx. But I still think I'm missing some Prefect configuration. For the client service in 
docker-compose
I set:
Copy code
- PREFECT__SERVER__ENDPOINT=<http://apollo:4200>
      - PREFECT__SERVER__UI__ENDPOINT=<http://ui:8080>
and when the flow is registred it's registred at:
Copy code
<http://ui:8080/default/flow/1dbc36f7-4302-458c-a8cd-e436a93142cb>
but from the UI I cannot access it. My question is, what URL should I set. Should I set the IP of the remote machine? @Ryan Peden What do you think?
5 Views
Open in Slack
PreviousNext
Powered by

Prefect Community

Bring your towel and join one of the fastest growing data communities. Welcome to our second-generation open source orchestration platform, a completely rethought approach to dataflow automation.

Powered by