R
RunPod9mo ago
Mihály

Failed to return job results.

Hi! I'm getting this error recenltly. I'm using runpod 1.6.2 {"requestId": "cce0d888-3d0c-4186-87cb-b94bfd359a71-e1", "message": "Failed to return job results. | 400, message='Bad Request', url=URL('https://api.runpod.ai/v2/8x6rjph4tvc8mp/job-done/hz8pr8qphlh72y/cce0d888-3d0c-4186-87cb-b94bfd359a71-e1?gpu=NVIDIA+GeForce+RTX+4090&isStream=false')", "level": "ERROR"} I'm not returning an error dict in the code. @ashleyk Could that be a sign that the result is over the size limit? Has the response size limit changed?
Solution:
Update For later readers with similar problem: In this specific case, i have found the root of the issue. It was indeed the response size, however it was not exceeded when returning the final response, but when using the runpod.serverless.progress_update ...
Jump to solution
5 Replies
ashleyk
ashleyk9mo ago
Which serverless worker is this? And what is the size of the response?
zyyy
zyyy9mo ago
Hi, may I know how to increase the response size ? Because I encountered this issue too when I upload a bigger image and the returned image is getting this error. However, it works well on small image size. Any solution for it ?
Mihály
MihályOP9mo ago
@ashleyk This is an automatic1111 worker. Whenever the size of the output array gets over 20MB, i'm encountering this. I guess then its the response size limit. The more baffling thing is that i'm getting this message also before the inference finishes, but only in case the image is big enough to reach 20MB. When this gets triggered, (underlined with red), then the /status check sends a response with
status:COMPLETED
status:COMPLETED
Any idea what could happen there?
No description
Mihály
MihályOP9mo ago
ps: Also, i'm uploading the results in code, and cleaning the response of them to avoid this, but still.
Solution
Mihály
Mihály9mo ago
Update For later readers with similar problem: In this specific case, i have found the root of the issue. It was indeed the response size, however it was not exceeded when returning the final response, but when using the runpod.serverless.progress_update Also, in an email conversation, the following was pointer out: - "limit for /runsync is 20mb" - "/run has an even smaller limit for 10mb at the moment" - "its both the input and the output payload limit"
Want results from more Discord servers?
Add your server