depot/third_party/nixpkgs/nixos/modules/services/misc/bees.nix
Default email 8ac5e011d6 Project import generated by Copybara.
GitOrigin-RevId: 2c3273caa153ee8eb5786bc8141b85b859e7efd7
2020-04-24 19:36:52 -04:00

123 lines
4.1 KiB
Nix

{ config, lib, pkgs, ... }:
with lib;
let
cfg = config.services.beesd;
logLevels = { emerg = 0; alert = 1; crit = 2; err = 3; warning = 4; notice = 5; info = 6; debug = 7; };
fsOptions = with types; {
options.spec = mkOption {
type = str;
description = ''
Description of how to identify the filesystem to be duplicated by this
instance of bees. Note that deduplication crosses subvolumes; one must
not configure multiple instances for subvolumes of the same filesystem
(or block devices which are part of the same filesystem), but only for
completely independent btrfs filesystems.
</para>
<para>
This must be in a format usable by findmnt; that could be a key=value
pair, or a bare path to a mount point.
'';
example = "LABEL=MyBulkDataDrive";
};
options.hashTableSizeMB = mkOption {
type = types.addCheck types.int (n: mod n 16 == 0);
default = 1024; # 1GB; default from upstream beesd script
description = ''
Hash table size in MB; must be a multiple of 16.
</para>
<para>
A larger ratio of index size to storage size means smaller blocks of
duplicate content are recognized.
</para>
<para>
If you have 1TB of data, a 4GB hash table (which is to say, a value of
4096) will permit 4KB extents (the smallest possible size) to be
recognized, whereas a value of 1024 -- creating a 1GB hash table --
will recognize only aligned duplicate blocks of 16KB.
'';
};
options.verbosity = mkOption {
type = types.enum (attrNames logLevels ++ attrValues logLevels);
apply = v: if isString v then logLevels.${v} else v;
default = "info";
description = "Log verbosity (syslog keyword/level).";
};
options.workDir = mkOption {
type = str;
default = ".beeshome";
description = ''
Name (relative to the root of the filesystem) of the subvolume where
the hash table will be stored.
'';
};
options.extraOptions = mkOption {
type = listOf str;
default = [];
description = ''
Extra command-line options passed to the daemon. See upstream bees documentation.
'';
example = literalExample ''
[ "--thread-count" "4" ]
'';
};
};
in {
options.services.beesd = {
filesystems = mkOption {
type = with types; attrsOf (submodule fsOptions);
description = "BTRFS filesystems to run block-level deduplication on.";
default = { };
example = literalExample ''
{
root = {
spec = "LABEL=root";
hashTableSizeMB = 2048;
verbosity = "crit";
extraOptions = [ "--loadavg-target" "5.0" ];
};
}
'';
};
};
config = {
systemd.services = mapAttrs' (name: fs: nameValuePair "beesd@${name}" {
description = "Block-level BTRFS deduplication for %i";
after = [ "sysinit.target" ];
serviceConfig = let
configOpts = [
fs.spec
"verbosity=${toString fs.verbosity}"
"idxSizeMB=${toString fs.hashTableSizeMB}"
"workDir=${fs.workDir}"
];
configOptsStr = escapeShellArgs configOpts;
in {
# Values from https://github.com/Zygo/bees/blob/v0.6.1/scripts/beesd%40.service.in
ExecStart = "${pkgs.bees}/bin/bees-service-wrapper run ${configOptsStr} -- --no-timestamps ${escapeShellArgs fs.extraOptions}";
ExecStopPost = "${pkgs.bees}/bin/bees-service-wrapper cleanup ${configOptsStr}";
CPUAccounting = true;
CPUWeight = 12;
IOSchedulingClass = "idle";
IOSchedulingPriority = 7;
IOWeight = 10;
KillMode = "control-group";
KillSignal = "SIGTERM";
MemoryAccounting = true;
Nice = 19;
Restart = "on-abnormal";
StartupCPUWeight = 25;
StartupIOWeight = 25;
SyslogIdentifier = "bees"; # would otherwise be "bees-service-wrapper"
};
wantedBy = ["multi-user.target"];
}) cfg.filesystems;
};
}