You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Currently, borg seems to list files to STDERR, which prevents its use in UNIX pipes. Of course, I could redirect 2>&1 but it would also include error messages and password prompts. The primary purpose of borg create --list --dry-run ... is to list files, so one could argue that therefore this should go to STDOUT and not STDERR. Also, this allows to use the borg pattern matching, then processing these somehow and finally pipe them back to borg create ... --paths-from-stdin.
My use-case is the following: I am using the pcloud remote of rclone to back up my data. This is slow when many small files are involved. As a test I arbitrarily tried to upload the git repo of opendataeditor and saw from vnstat -l that upload speed is only a few kbit/s. As it was noted that multiple parallel borg2 processes can concurrently write to the same repo, I did exactly that.
However, these processes tended to "clump together" uploading the same files at the same time, which is why I wanted to shuffle the file order so that the different upload processes do not interfere as much. This is where borg create --list --dry-run ... | shuf would become handy.
The text was updated successfully, but these errors were encountered:
Currently, borg seems to list files to STDERR, which prevents its use in UNIX pipes. Of course, I could redirect
2>&1
but it would also include error messages and password prompts. The primary purpose ofborg create --list --dry-run ...
is to list files, so one could argue that therefore this should go to STDOUT and not STDERR. Also, this allows to use the borg pattern matching, then processing these somehow and finally pipe them back toborg create ... --paths-from-stdin
.My use-case is the following: I am using the pcloud remote of rclone to back up my data. This is slow when many small files are involved. As a test I arbitrarily tried to upload the git repo of opendataeditor and saw from
vnstat -l
that upload speed is only a few kbit/s. As it was noted that multiple parallel borg2 processes can concurrently write to the same repo, I did exactly that.However, these processes tended to "clump together" uploading the same files at the same time, which is why I wanted to shuffle the file order so that the different upload processes do not interfere as much. This is where
borg create --list --dry-run ... | shuf
would become handy.The text was updated successfully, but these errors were encountered: