Hello, I am testing prefect setup using KIND. Basically there is no difference between kind and kubernetes. I am facing to connect to graphql api. Have anybody had similar problem?

Hi @Rafal - can you confirm the API is available at that address and port from elsewhere?

so, after visiting I can see statement

GET query missing

Ok great, that's what you'd expect if it was available

That's right

Hm that's interesting. Are you able to remove the existing UI image and spin up server again? (or just the UI image at least)

ok. I deployed UI once again. The same result 🤔

That's very odd. Is that endpoint open to all external traffic? Or are there some VPC rules attached?

What do you mean by saying extarnal? There are no VPC rules behind it, in kind kubernetes nodes are simulated by containerds.

Ah I hadn't noticed that was a 172 IP, my fault. I think I see the issue though; try adding the protocol to your config (

<http://172.18.0.2:32000/graphql>

) and then removing the UI image and restarting

I've already added it in UI deployment

env:

- name: PREFECT_SERVER__GRAPHQL_URL

value: "172.18.0.2:32000/graphql"

Yes you're missing the protocol

http

@nicholas now all is green! Seems like working! But there is another question: how to register flow? I guess it comes to start agent... 🤔

To register a flow, you'll want to add this to your local Prefect `~/.prefect/config.toml`:

[server]
endpoint = "172.18.0.2:32000/graphql"

Ok! I've added

- name: PREFECT_SERVER__ENDPOINT

value: "172.18.0.2:32000/graphql"

to prefecthq/server image. Error:

requests.exceptions.ConnectionError: HTTPConnectionPool(host='localhost', port=4200): Max retries exceeded with url: /graphql/alpha (Caused by NewConnectionError('<urllib3.connection.HTTPConnection object at 0x7ff77c2d35b0>: Failed to establish a new connection: [Errno 111] Connection refused')

🤔

When's that error happening @Rafal?

No, should I? When I am trying to register flow. Do I miss something? How to register flow when prefect is running on "local" cluster?

No, just checking. The endpoint variable needs to be set in the environment you're registering flows from, i.e. your local machine. You can do so easily with the

~/.prefect/config.toml

configuration file.

Ok, but all of prefect's components are running on cluster. What should run on my localhost excactly? I need some port forwarding? I do not use confi.toml, just notation from docs:

PREFECT__ and two underscores (__)

Understood, when you're writing/registering flows on your computer, Prefect needs to know where your server is running, if it's not localhost or Cloud

So. maybe one of the solution is to add volume with flows to

prefecthq/server

image and register it directly from there as localhost? Is there another option besides Cloud solution?

There's no need for that, you just need to set the local config that I mentioned, then whenever you register a flow locally, it will register with your cluster

OK! So I needed to create

~/.prefect/config.toml

and after providing config which you suggested I've got:

raise InvalidSchema("No connection adapters were found for {!r}".format(url))
requests.exceptions.InvalidSchema: No connection adapters were found for '172.18.0.2:32000/graphql/graphql/alpha'

Which version of Prefect Core do you have locally? I think updating your version should fix that

0.11.5

I will try with update

Hm let me check then

@nicholas next step closer to solution 😉 It was registered, but two things wich I am worry about:

Flow: <http://localhost:8080/flow/18738ebf-fd53-4831-9157-fbb384cc6f3c>

- still localhost ? Second one: tasks can't run. Some strange monit about Late runs. Does it mean that scheduler is not working properly?

I think setting

[server]
host = "<http://172.18.0.2>"

will fix that flow.register output but it's not impacting the registration at all. Do you have an agent running? That'll be needed to pick up runs.

@nicholas prefect works perfectly fine! Seems like agent need to run all the time to keep flows up to date. Should I keep agent running in container on cluster for example? Sometimes I need to turn off my local computer 😉

Yes, an agent will need to be running to pick up any flow runs; I'd recommend taking a look at https://docs.prefect.io/orchestration/tutorial/multiple.html#install-a-supervised-agent for some supervised agent info