Websocket Runpod proxy problems
When uploading a 1.8 MiB file over a websocket connection to the pod via the runpod proxy, the websocket is force-disconnected. It's not an error related to the software, the disconnect appears to come from the proxy.
It might be a case where the server autoterminate the connection if it's inactive for X many seconds. So the limit on filesize is however much data I can manage to upload before it treats the connection as inactive?
Anything that can be done to fix this?
...
A bit more details on my specific situation if needed:
I'm running SwarmUI and FLUX to generate images and using the "Init Image" (img2img) feature a lot.
If I use a very small file like 8kb image there's no issues.
But even when I just start using an 1.8MiB generated image, I get these errors a lot.
Something like 20%+ of the times. And then it might happen the next 1-10 times I click generate a new image, and then suddenly it works anyway. A bit weird.
It happens on many different GPUs in many different regions (all in the secure cloud) also on different days.
I access SwarmUI through this template "SwarmUI Jupyter -latest" which opens a new window like
https://8m6og5cjkl363u-2254.proxy.runpod.net/Text2Image
My internet connection is stable.
I've talked to both the maintainer of SwarmUI and the guy who created the template. They both point me to ask you guys.
(Just FYI this issue also happens on Massedcompute when I launch SwarmUI in some sort of temporary Cloudflare link there)
It might be a case where the server autoterminate the connection if it's inactive for X many seconds. So the limit on filesize is however much data I can manage to upload before it treats the connection as inactive?
Anything that can be done to fix this?
...
A bit more details on my specific situation if needed:
I'm running SwarmUI and FLUX to generate images and using the "Init Image" (img2img) feature a lot.
If I use a very small file like 8kb image there's no issues.
But even when I just start using an 1.8MiB generated image, I get these errors a lot.
Something like 20%+ of the times. And then it might happen the next 1-10 times I click generate a new image, and then suddenly it works anyway. A bit weird.
It happens on many different GPUs in many different regions (all in the secure cloud) also on different days.
I access SwarmUI through this template "SwarmUI Jupyter -latest" which opens a new window like
https://8m6og5cjkl363u-2254.proxy.runpod.net/Text2Image
My internet connection is stable.
I've talked to both the maintainer of SwarmUI and the guy who created the template. They both point me to ask you guys.
(Just FYI this issue also happens on Massedcompute when I launch SwarmUI in some sort of temporary Cloudflare link there)
