pyrox.dev
1{ config, lib, pkgs, ... }:
2
3with lib;
4
5let
6
7 cfg = config.services.beesd;
8
9 logLevels = { emerg = 0; alert = 1; crit = 2; err = 3; warning = 4; notice = 5; info = 6; debug = 7; };
10
11 fsOptions = with types; {
12 options.spec = mkOption {
13 type = str;
14 description = lib.mdDoc ''
15 Description of how to identify the filesystem to be duplicated by this
16 instance of bees. Note that deduplication crosses subvolumes; one must
17 not configure multiple instances for subvolumes of the same filesystem
18 (or block devices which are part of the same filesystem), but only for
19 completely independent btrfs filesystems.
20
21 This must be in a format usable by findmnt; that could be a key=value
22 pair, or a bare path to a mount point.
23 Using bare paths will allow systemd to start the beesd service only
24 after mounting the associated path.
25 '';
26 example = "LABEL=MyBulkDataDrive";
27 };
28 options.hashTableSizeMB = mkOption {
29 type = types.addCheck types.int (n: mod n 16 == 0);
30 default = 1024; # 1GB; default from upstream beesd script
31 description = lib.mdDoc ''
32 Hash table size in MB; must be a multiple of 16.
33
34 A larger ratio of index size to storage size means smaller blocks of
35 duplicate content are recognized.
36
37 If you have 1TB of data, a 4GB hash table (which is to say, a value of
38 4096) will permit 4KB extents (the smallest possible size) to be
39 recognized, whereas a value of 1024 -- creating a 1GB hash table --
40 will recognize only aligned duplicate blocks of 16KB.
41 '';
42 };
43 options.verbosity = mkOption {
44 type = types.enum (attrNames logLevels ++ attrValues logLevels);
45 apply = v: if isString v then logLevels.${v} else v;
46 default = "info";
47 description = lib.mdDoc "Log verbosity (syslog keyword/level).";
48 };
49 options.workDir = mkOption {
50 type = str;
51 default = ".beeshome";
52 description = lib.mdDoc ''
53 Name (relative to the root of the filesystem) of the subvolume where
54 the hash table will be stored.
55 '';
56 };
57 options.extraOptions = mkOption {
58 type = listOf str;
59 default = [ ];
60 description = lib.mdDoc ''
61 Extra command-line options passed to the daemon. See upstream bees documentation.
62 '';
63 example = literalExpression ''
64 [ "--thread-count" "4" ]
65 '';
66 };
67 };
68
69in
70{
71
72 options.services.beesd = {
73 filesystems = mkOption {
74 type = with types; attrsOf (submodule fsOptions);
75 description = lib.mdDoc "BTRFS filesystems to run block-level deduplication on.";
76 default = { };
77 example = literalExpression ''
78 {
79 root = {
80 spec = "LABEL=root";
81 hashTableSizeMB = 2048;
82 verbosity = "crit";
83 extraOptions = [ "--loadavg-target" "5.0" ];
84 };
85 }
86 '';
87 };
88 };
89 config = {
90 systemd.services = mapAttrs'
91 (name: fs: nameValuePair "beesd@${name}" {
92 description = "Block-level BTRFS deduplication for %i";
93 after = [ "sysinit.target" ];
94
95 serviceConfig =
96 let
97 configOpts = [
98 fs.spec
99 "verbosity=${toString fs.verbosity}"
100 "idxSizeMB=${toString fs.hashTableSizeMB}"
101 "workDir=${fs.workDir}"
102 ];
103 configOptsStr = escapeShellArgs configOpts;
104 in
105 {
106 # Values from https://github.com/Zygo/bees/blob/v0.6.5/scripts/beesd@.service.in
107 ExecStart = "${pkgs.bees}/bin/bees-service-wrapper run ${configOptsStr} -- --no-timestamps ${escapeShellArgs fs.extraOptions}";
108 ExecStopPost = "${pkgs.bees}/bin/bees-service-wrapper cleanup ${configOptsStr}";
109 CPUAccounting = true;
110 CPUSchedulingPolicy = "batch";
111 CPUWeight = 12;
112 IOSchedulingClass = "idle";
113 IOSchedulingPriority = 7;
114 IOWeight = 10;
115 KillMode = "control-group";
116 KillSignal = "SIGTERM";
117 MemoryAccounting = true;
118 Nice = 19;
119 Restart = "on-abnormal";
120 StartupCPUWeight = 25;
121 StartupIOWeight = 25;
122 SyslogIdentifier = "beesd"; # would otherwise be "bees-service-wrapper"
123 };
124 unitConfig.RequiresMountsFor = lib.mkIf (lib.hasPrefix "/" fs.spec) fs.spec;
125 wantedBy = [ "multi-user.target" ];
126 })
127 cfg.filesystems;
128 };
129}