TL;DR: Perma-seeding of whole Servarr libraries optimized for per-tracker ratio.
- Delete torrents/items only as disk space gets low.
- Don't delete currently imported items. IOW, only delete upgraded items.
- Don't delete private items that haven't met seeding requirements.
- Delete public items first.
- Delete private items in an order to maximize tracker ratio and/or bonuses.
- Delete stalled items , and items containing archives such as *.rar releases and blacklist them, AKA mark them as failed, in Servarr.
- And more...
Table of Contents
Seed Servarr download client torrents/items as long as possible only deleting them as necessary as disk space gets low, hence the name based on "to prune". Which download items are considered eligible for deletion is configured by the user. The common case is that download items that are currently imported are not considered for deletion. Neither are items from private trackers/indexers that have been upgraded or otherwise deleted from the library but haven't met the indexers seeding requirements. The order in which download items are deleted is determined according to rules configured by the user. The common case is to delete items from public indexers first and among those to delete the items with the highest ratio first to preserve the health of the community by seeding less popular items longer. Next delete items from private indexers by configured indexer priority and within the items for a given indexer to delete items in an order to maximize ratio and/or seeding rewards.
Other configured operations may be applied as well. For example:
- Verify and resume corrupt items
- Increase bandwidth priority for items from private indexers
- Decrease bandwidth priority for items from public indexers
- Remove and blacklist download items containing archives (
*.rar
,*.zip
,*.tar.gz
, etc.) which can't be perma-seeded - Remove and blacklist stalled download items
- etc.
The $ prunerr
command is intended to serve as a companion to the Servarr [1] suite of
applications and services and the Transmission BitTorrent client [2]. It periodically
polls the download clients [3] of Sonarr [4], Radarr [5], etc. and applies the configured
operations to the download items in each of those download clients. It can also be run
independently of any Servarr instances to optimize seeding for download items added by
other means, e.g. FlexGet [6].
See the Usage section below for full details.
Install locally or use the Docker container image:
Install by using any tool for installing standard Python 3 distributions such as pip [7]:
$ pip3 install --user prunerr
Optional shell prompt tab completion is available by using argcomplete [8].
The recommended way to use the container image is by using Docker Compose [9]. See the example ./docker-compose.yml file [10]. Write your configuration and run the container:
$ docker compose up
You can also use the image directly. Pull the Docker image [11]. Use it to create and run a container:
$ docker pull "registry.gitlab.com/rpatterson/prunerr" $ docker run --rm -it "registry.gitlab.com/rpatterson/prunerr" ...
Use image variant tags to control when the image updates. Releases publish tags for the
branch and for major and minor versions. For example, to keep up to date with a specific
branch, use a tag such as
registry.gitlab.com/rpatterson/prunerr:main
. Releases from develop
publish pre-releases. Releases from main
publish final releases. Releases from
main
also publish tags without a branch, for example
registry.gitlab.com/rpatterson/prunerr
. Releases from main
also
publish tags for the major and minor version, for example
registry.gitlab.com/rpatterson/prunerr:v0.8
.
Releases publish multi-platform images for the following platforms:
linux/amd64
linux/arm64
linux/arm/v7
Start by writing your ~/.config/prunerr.yml
configuration file. See the comments in
the example configuration [12] for details.
Once configured, you may run individual sub-commands once, run all operations once as
configured using the $ prunerr exec
sub-command, or run all operations in a polling
loop using the $ prunerr daemon
sub-command. See the Order of Operations section
for a detailed description of the operations. Use the CLI help to list the other
sub-commands and to get help on the individual sub-commands:
$ prunerr --help $ prunerr exec --help
If using the Docker container image, the container can be run from the command-line as well:
$ docker compose run "prunerr" prunerr --help
Note that polling is required because there is no event we can subscribe to that
reliably determines disk space margin as the download clients are downloading. Every
run of the $ prunerr exec
sub-command or every loop of the $ prunerr daemon
sub-command performs the following operations.
Verify and resume corrupt items, same as:
$ prunerr verify
.Review download items, same as:
$ prunerr review
:Apply per-indexer review operations as configured under
indexers/reviews
in the configuration file to all download items.Move download items that have been acted on by Servarr to the
*/seeding/*
directory, same as:$ prunerr move
.As Servarr acts on completed download items, be that importing files from them, ignoring them, deleting them from the queue, etc., Prunerr moves those items from the Servarr download client's
Directory
to a parallel*/seeding/*
directory. Then when deleting download items to free space, Prunerr only considers items under that directory. This has the added benefit of reflecting which items have been acted on by Servarr in the download client.Delete download items if disk space is low, same as:
$ prunerr free-space
.Consider items for deletion in different groups in this order:
Download items no longer registered with tracker.
IOW, items that can no longer be seeded at all first.
Orphan files and directories not belonging to any download item
Walk all the top-level directories used by each download client and identify which paths don't correspond to a download client item.
Warning
Stop Prunerr before moving any download items.
Prunerr uses the list of files for every download item to identify orphans. If a user, or anything else other than Prunerr, moves or changes the location of a download item while Prunerr is identifying orphans, Prunerr may identify the new download item data path as an orphan and delete it out from under the download item leading to data loss.
Imported/seeding download items
IOW, download items that have been acted upon by Servarr and moved to the
*/seeding/*
directory by the$ prunerr move
sub-command/operation excluding those items filtered out according to theindexers/priorities
operations withfilter: true
. For example, don't delete currently imported items (by hard link count) or items that haven't met private indexer seeding requirements.
For each of these groups in order, loop through each item in the group and:
- Check disk space against the margin configured by
download-clients/*/max-download-bandwidth
anddownload-clients/*/min-download-time-margin
- If there's insufficient disk space, delete the item.
For the orphans group, delete smaller items first to minimize the amount of re-downloading needed should the user notice and correct any issues resulting in the orphans.
For the other groups delete items in the order determined by the configured
indexers/priorities
indexer order then by the configured operations for that item's indexer.For those times when there's nothing Prunerr can delete to free disk space, most users' download clients should also probably pause downloading when disk space drops significantly below this margin. Use the provided Transmission pause download script [13] and, optionally, integrate it into your Docker Compose project via a cron job [14] or see those as examples.
The export
sub-command is roughly the inverse of Servarr import events, hard link
imported files back into download client items and verify. See the CLI $ prunerr
export --help
output for more details.
GitLab hosts this project [15] and mirrors it to GitHub [16] but use GitLab for reporting issues, submitting pull or merge requests and any other development or maintenance activity. See the contributing documentation [17] for more details on how to get started with development.
I didn't like the available options I could find at the time for maximizing seeding from a lovingly managed media library. Deleting by a ratio threshold doesn't make sense to me because that can delete items when there's plenty of disk space. Also the ratio threshold is a reverse indicator for items from private indexers vs items from public indexers. Items from private indexers with high ratios should be kept around as long as possible to build user total ratio whereas items from public indexers with low ratios should be kept around as long as possible to preserve access in the community/ecosystem. Finally, deleting any item still imported in the Servarr only because it hit the ratio threshold is the biggest waste since it doesn't free any space. So I wrote Prunerr to prune download items in the correct order.
The use case for Prunerr is not tracker ratio racing. It's goal is to seed as long as possible and to seed as much of your library as possible. This should have some secondary benefits to ratio, but that's not the main goal.
Finally, there is a laundry list of other download client management tasks that can be automated but aren't by anything I could find. So I added them to Prunerr as well.
[1] | https://wiki.servarr.com |
[2] | https://transmissionbt.com/ |
[3] | https://wiki.servarr.com/radarr/settings#download-clients |
[4] | https://wiki.servarr.com/en/sonarr |
[5] | https://wiki.servarr.com/en/radarr |
[6] | https://flexget.com/ |
[7] | https://pip.pypa.io/en/stable/installation/ |
[8] | https://kislyuk.github.io/argcomplete/#installation |
[9] | https://docs.docker.com/compose/ |
[10] | https://gitlab.com/rpatterson/prunerr/-/blob/main/docker-compose.yml |
[11] | https://hub.docker.com/r/merpatterson/prunerr |
[12] | https://gitlab.com/rpatterson/prunerr/blob/main/src/prunerr/home/.config/prunerr.yml |
[13] | https://gitlab.com/rpatterson/prunerr/blob/main/transmission/usr/local/bin/transmission-pause-download |
[14] | https://gitlab.com/rpatterson/prunerr/blob/main/transmission/etc/crontabs/abc |
[15] | https://gitlab.com/rpatterson/prunerr |
[16] | https://github.com/rpatterson/prunerr |
[17] | https://gitlab.com/rpatterson/prunerr/-/blob/main/docs/contributing.rst |