Open quetil opened 3 years ago
Hello Adrien!
We are using gRPC for tensorboard dev
calls so perhaps the support you need is already built in at lower levels. Could you try some of the environment variables described here?
https://github.com/grpc/grpc/blob/master/doc/environment_variables.md
I haven't really tried it but perhaps https_proxy
and GRPC_DEFAULT_SSL_ROOTS_FILE_PATH
are the variables you need.
If you get it to work, do you mind posting some details here? If you're having troubles please feel free to post followup questions.
Hello Brian,
Thanks for the reply, very much appreciated ! Well searched, I thought it was going to work ! But still the following error message:
Failed to connect to backend: HTTPSConnectionPool(host='tensorboard.dev', port=443): Max retries exceeded with url: /api/uploader (Caused by SSLError(SSLError(1, '[SSL: WRONG_VERSION_NUMBER] wrong version number (_ssl.c:1124)')))
It seems grpc doesn't look at the GRPC_DEFAULT_SSL_ROOTS_FILE_PATH 🤔
I'll continue to investigate and keep this thread in touch if any progress!
Sorry, there was a typo in my response, which I have updated. The name of the env var is actually 'GRPC_DEFAULT_SSL_ROOTS_FILE_PATH'.
I have been able to configure GRPC with this variable in the past. I've just never done it in combination with a proxy.
No worries, I noticed the typo and already tried with the right env var GRPC_DEFAULT_SSL_ROOTS_FILE_PATH
, but still the same error message. Maybe it's not the right pem file (but it's working for pip and conda for instance).
I'll continue to investigate
@quetil , can this issue be closed?
I'd prefer to be sure and to have someone which is telling us "it's working" in a classic environnement. And then, if it's the case, it will probably be a wrong config in the enterprise virtual machine I'm working on, which is blocking the right behavior.
Hello all,
I'd like to test and use
tensorboard dev
, but I'm working behind a proxy. So when I write the proxy in the environment variable, I got a SLL error (or certificate chain error). So I looked for a way to write the cafile.pem location so tensorboard can use it, but I didn't find a way. Usually, for such command there is a flag--cafile
and a flag--proxy
, so maybe it's feature that can be requested?Do you have any ideas?
Thank you in advance,
Adrien