S3 backup fails after a few minutes

[jaw12346]

I just upgraded in-place from PBS 3 to PBS 4, and I'm trying to set up S3 backups for my vital CTs. I've been able to back up ~1gb CTs, but when I tried a 20gb VM and a 250+gb CT, it runs for 3-6 minutes (usually around 3.5 mins) and then fails with the following error every time:

2025-10-17T16:27:08-04:00: POST /fixed_chunk: 400 Bad Request: failed to upload chunk to s3 backend - upload failed: client error (SendRequest): connection error: Connection reset by peer (os error 104) 2025-10-17T16:27:08-04:00: backup ended and finish failed: backup ended but finished state is not set. 2025-10-17T16:27:08-04:00: removing unfinished backup 2025-10-17T16:27:09-04:00: removing backup snapshot "/s3-cache/vm/107/2025-10-17T20:25:20Z" 2025-10-17T16:27:09-04:00: TASK ERROR: backup ended but finished state is not set.

I'm using Backblaze B2 (not free-tier). Since I'm still experimenting, against the guidance of the docs, my cache is set up on the PBS root disk and doesn't have any capacity restrictions; however, only around 10% of the disk is in use. My PVE is still on 8.4.14.

Please share if you have had similar experiences and have found a workaround.

 

[Chris]

Hi,

400 Bad Request: failed to upload chunk to s3 backend - upload failed: client error (SendRequest): connection error: Connection reset by peer (os error 104)

seems like the provider api endpoint closes the connection while your PBS is sending the upload requests. Most likely because of traffic limiting.

Please try and add a put-rate-limit to your S3 endpoint configuration as described in https://forum.proxmox.com/threads/s3-upload-fails-with-large-vms-cache-issue.169432/post-792910, maybe starting with the lowest vaule of 1 ( = 1 request/second ) and see if the issue persists.

 

[jaw12346]

@Chris I think you're right about it being a traffic limiting issue. I made an account with Wasabi and uploads have been working for the past few days.