failing to start job
One of 10 times we are getting error message when trying to pass message. The error is no inside serverless container. Job is not getting processed by runpod itself.
running from fastapi background_task. Sot the trace is not full

11 Replies
"Connection Reset", are you still having this issue?
Sorry, I didn't get your point? We are doing retrials and it helps. But I hope the serverless functionality is not intended to be so unstable
I was pointing out the error and asking if you were still seeing this issue. I'll take a look into it for you, this isn't normal activity and can happen for a lot of reasons.
is there something blocking your connection to runpod? vpn's, blocking from country to external sites?
and how big is your payload in mb's?
The payload is small, something like couple kb's. We are not sending any media in the payload.
As for the blocking the connection. Up to my knowledge, there is no blocks from the country side. Is there a possibility of blocking on the runpod side?
The theory about blocking is kinda weak, as we have the same static ip and requests fail sometimes, but not always
hmm yeah im just guessing, its best to open a ticket and let staff's handle these
any endpoint id's?
Do you normally have connection issues to Cloudflare? Your connection through the proxy is a Cloudflare Tunnel :thinkMan:
Another user saw success switching to a TCP port and their own CF Tunnel iirc
I am sorry for the late response. I just checked, Clodflare services are available. What do you mean by "Your connection through the proxy"? I am not using proxy
Yo be precise, we are using Yandex Cloud as hosting provider
I asked the our provider's assistance. They recommended to debug the connection using the following command. What port and ip should I set?
mtr -rzbw --tcp --port <port> -c 20 <address>
Wait this is for endpoint serverless tho, they shouldnt be using tunnels
I think you can't do this to send a job
Any suggestions on debugging?
Oh it still throws that error?
if you try it using a vps from a known vps provider like digitalocean/vultr/aws/azure/google cloud, des it work?
did you run it from your local pc?
maybe you should open a support ticket