Output of restic version

Built from master on 29th of July:

restic 0.17.0-dev (compiled manually) compiled with go1.22.0 on linux/amd64

What backend/service did you use to store the repository?

s3 (served via standalone minio deployment)

Problem description / Steps to reproduce

Getting "context canceled" while doing prune operations lately.

Expected behavior

Operations should have a chance to finish, even on extremely slow endpoints.

Actual behavior

Depending on the IO status on the hardware (which is a busy one with multiple IO-heavy processes are running in parallel, albeit sometimes slow), some operations gets canceled. E.g.

List(data) returned error, retrying after 1s: Get "http://myhost:8888/mybucket/?continuation-token=ZGF0YS83Ni83NmQ4OTg2MzUzOGMxMjMxOTU2MDI2Mzg2M2M4NmM0ZTM3NzNjMjU2Y2I2MjBiZWJjYzVhZDQzY2E3ZWQ0NjIwW21pbmlvX2NhY2hlOnYyLGlkOmFiZjQ5YjFkLTVjYTItNDY3MS1hNzM4LWIyMjFkZDIwNTMzOSxwOjAsczowXQ%3D%3D&delimiter=&encoding-type=url&fetch-owner=true&list-type=2&prefix=data%2F": context canceled

Do you have any idea what may have caused this?

A while ago I asked the question on the forum and turned out changing one value was solving the issue (or disabling the backend-error-redesign feature flag).

Wanted to open this as an issue to make it traceable, since this is a blocking thing on slow endpoints. Plus the flag has been enabled as beta on release version also.

Even though the endpoint is a lot slower than expected, I don't think restic should require a "minimum speed" from endpoints. Maybe the timeout can be configured with a special argument (sorry if there is already one, I could not find digging the code).

Did restic help you today? Did it make you happy in any way?

Still the best! 🚀