Restic panics on connection reset when backing up to S3

[lx7]

Output of restic version

restic 0.2.0 (v0.2.0-0-gc7f5ac2)
compiled at 2016-07-31 10:26:08 with go1.6.2

Expected behavior

Restic should be able to handle connection issues gracefully and retry. If after several retries the connection can not be re-established, restic should abort properly with a defined exit code.

Actual behavior

The restic process crashes with an unhandled panic() on disconnect.

scan [/srv/data/pictures]
scanned 21389 directories, 106521 files in 0:24
panic: Put https://my-bucket-name.s3-eu-west-1.amazonaws.com/restic/data/8b613a8128467e097483262eadaefd99a0763449cb09d0d367390cdf7ea69031: read tcp 192.168.178.2:24009->54.231.131.0:
443: read: connection reset by peer

goroutine 90776 [running]:
panic(0x9cb760, 0xc8234d0240)
        /usr/local/go/src/runtime/panic.go:481 +0x3e6
restic.(*Archiver).saveChunk(0xc821b26280, 0x1afa0b7, 0x2d2427, 0x512a07a100000, 0xc8400fc000, 0x2d2427, 0x4b6000, 0xc82387b600, 0x800e86520, 0xc8201487a8, ...)
        /tmp/restic-build-339571701/src/restic/archiver.go:165 +0x12b
created by restic.(*Archiver).SaveFile
        /tmp/restic-build-339571701/src/restic/archiver.go:237 +0x556

There's actually a TODO for this case at archiver.go:165

// TODO handle error
if err != nil {
    panic(err)
}

Steps to reproduce the behavior

Disconnect when backing up to S3.

[fd0]

Hi, thank for creating an issue about this. It sounds similar to #415, although we've updated the minio-go library, which should handle reconnect to s3 internally.

To solve this problem, I'd like to implement a retry logic independently of the backend (see also #353).

[fd0]

This should be solved with the latest update of the s3 connection library we're using. Please try again. Until then I'll close this report, we can always reopen it later.