Possible bug with ssh/sftp implementation - may try to delete failed uploads on a broken connection after timeout

[scotte]

It's hard to know with 100% certainty that this is the cause, but it seems like after a timeout in uploading via ssh/sftp that it then tries to delete the failed upload on the broken connection resulting in error that the remote file doesn't exist - when it actually does exist.

# zfsbackup send --compressor zfs --separator '+' zroot/dataset@autosnap_2025-02-01_00:00:07_monthly ssh://redacted/path/to/remote
2025/02/14 10:26:13 ssh backend: Error closing volume zroot/dataset+autosnap_2025-02-01_00:00:07_monthly.zstream.vol1 - connection lost
2025/02/14 10:26:13 ssh backend: Error deleting failed upload file /path/to/remote/zroot/dataset+autosnap_2025-02-01_00:00:07_monthly.zstream.vol1 - remove /path/to/remote/zroot/dataset+autosnap_2025-02-01_00:00:07_monthly.zstream.vol1: no such file or directory

$ ssh redacted ls -lhR /path/to/remote
total 141
drwxr-xr-x  2 xxx  yyy     4B Feb 14 17:03 zroot

/path/to/remote:
total 357145
-rw-r--r--  1 xxx  yyy   184M Feb 14 17:09 dataset+autosnap_2025-02-01_00:00:07_monthly.zstream.vol1