Failed to connect to all addresses

[knorr3]

Hi,
I got the server running and tried it out. After executing the client:

doc = Document(text=prompt).post(server_url, parameters={'num_images': 8})
da = doc.matches

The server doesn't respond to the notebook and prints the following error:

DEBUG  diffusion/rep-0@24 recv DataRequest at / with id:     [08/11/22 06:41:47]
       4856d9e7b3334adf9aec568629267f62                                         
DEBUG  dalle/rep-0@23 recv DataRequest at / with id:         [08/11/22 06:41:47]
       4856d9e7b3334adf9aec568629267f62                                         
INFO   diffusion/rep-0@24 diffusion  ...                     [08/11/22 06:41:47]
INFO   dalle/rep-0@23 dalle 8 ...                            [08/11/22 06:41:47]
DEBUG  diffusion/rep-0@24 connected to demo-cas.jina.ai:2096 [08/11/22 06:41:47]
ERROR  diffusion/rep-0@24 gRPC error: StatusCode.UNAVAILABLE [08/11/22 06:42:27]
       failed to connect to all addresses                                       
       The ongoing request is terminated as the server is                       
       not available or closed already.                                         
ERROR  diffusion/rep-0@24 ConnectionError('failed to connect [08/11/22 06:42:27]
       to all addresses')                                                       
        add "--quiet-error" to suppress the exception                           
       details

Any idea what I can do?

EDIT: I am running the latest docker image on OpenShift.

[derek-adair]

You have not configured your server_url properly would be my guess.

[delgermurun]

@knorr3 have you figured out the issue?

Maybe demo-cas.jina.ai:2096 is not reachable for some reason.
If you have extra GPU memory, you could also try to run CLIP yourself
https://github.com/jina-ai/dalle-flow/blob/main/README.md#run-your-own-clip

[knorr3]

Ah, now I get it! It cannot connect to demo-cas.jina.ai because it has to use the proxy.
The documentation of jina doesn't state any proxy settings for executors :(

I tried your suggestion, but as I am in a container in a Kubernetes cluster, i cannot run docker inside it.

[delgermurun]

It cannot connect to demo-cas.jina.ai because it has to use the proxy

Usually you don't need to do anything special if your machine has public internet access (outgoing traffic is allowed).

[knorr3]

I cannot access public internet without using the proxy. Is there any way I can configure jina to use the proxy?

[derek-adair]

Jina can be configured to point to a proxy from the CLI with the --proxy flag. I've not really used k8s but i'm assuming you can set that flag on some entrypoint-like-thing.

[knorr3]

It does have a --proxy flag, but only for jina gateway and jina client.
There is no proxy flag for the jina flow command:

jina flow [-h] [--name] [--workspace] [--log-config] [--quiet] [--quiet-error]
          [--optout-telemetry] [--uses] [--env [KEY: VALUE [KEY: VALUE ...]]]
          [--inspect {HANG, REMOVE, COLLECT}]

[delgermurun]

i cannot run docker inside it.

Or you can run CLIP executor from its source code.
Try running with uses: jinahub://CLIPTorchEncoder/latest-gpu? (instead of jinahub+docker)

I don't know about --proxy flag, sorry. You can always ask from support channel in Slack.

[knorr3]

I tried using jinahub:// but it just terminated after downloading, again without any error.

I think the most reasonable action at this point would be to get a proxy whitelisting for demo-cas.jina.ai.
Thank you very much for the support. I'll write again when there are any news or problems.