S3 connection resets

Getting Help
1

Hi forum ,

I’m evaluating restic at work to backup data to amazon S3. Backup works fine, but restore failed with messages : " read: connection reset by peer" to Amazon S3.

The process finish with a lot of missing files :

Summary: Restored 1807738 / 1812848 files/dirs (148.188 GiB / 148.205 GiB) in 40:13
Fatal: There were 348663 errors

Do you know if there is any way to debug or tune restic ?

Regards,

2

Sounds like an unstable connection or something? Have you tried it from other connections that are known to be stable?

See here for information on debugging: Participating — restic 0.16.2 documentation

3

Adjust restics retry settings --s3.connections, --s3.max-retries, --s3.retry-delay to enhance network resilience during restores from Amazon S3.
Example

restic restore --s3.connections 10 --s3.max-retries 20 --s3.retry-delay 5s <other-options>

4

@nehakakar Each option needs to be prefixed with -o or --option (see restic options), so e.g. -o s3.connections=10rather than–s3.connections 10`.

5

Hello, I have similar issues and tried your options.

mfs@hsd92:/data# restic restore 0ce15152 -o s3.connections=10 -o s3.max-retries=20 -o s3.retry-delay=5s  --target /data/tmp2/
Fatal: option s3.max-retries is not known

Next to the max-retries error, I do not even find the options on any doc page. How do you came up with those?

Thanks,
Meffesino

6

The current list of S3 options according to restic options:

  s3.bucket-lookup    bucket lookup style: 'auto', 'dns', or 'path'
  s3.connections      set a limit for the number of concurrent connections (default: 5)
  s3.layout           use this backend layout (default: auto-detect)
  s3.list-objects-v1  use deprecated V1 api for ListObjects calls
  s3.region           set region
  s3.retries          set the number of retries attempted
  s3.storage-class    set S3 storage class (STANDARD, STANDARD_IA, ONEZONE_IA, INTELLIGENT_TIERING or REDUCED_REDUNDANCY)