From fefc7bd20db473db1104150f2631b365ffaeb159 Mon Sep 17 00:00:00 2001
From: Katherina Walshe-Grey <git@qenya.tel>
Date: Tue, 4 Mar 2025 14:43:03 +0000
Subject: [PATCH] backup: init as a reusable module

---
 flake.nix                     |  15 ++++
 flake/backup.nix              | 126 ++++++++++++++++++++++++++++++++++
 flake/default.nix             |   1 +
 hosts/elucredassa/default.nix |  20 +-----
 hosts/orm/default.nix         |  13 ----
 5 files changed, 144 insertions(+), 31 deletions(-)
 create mode 100644 flake/backup.nix

diff --git a/flake.nix b/flake.nix
index 307a812..5562b16 100644
--- a/flake.nix
+++ b/flake.nix
@@ -102,6 +102,21 @@
         "qenya@shaw".imports = [ ./hosts/shaw/home.nix ];
       };
 
+      fountain.backup = {
+        keys = {
+          elucredassa = [ "ssh-ed25519 AAAAC3NzaC1lZDI1NTE5AAAAIOFa3hjej6KGmS2aQ4s46Y7U8pN4yyR2FuMofpHRwXNk syncoid@elucredassa" ];
+        };
+        sync = {
+          "orm-state" = {
+            dataset = "state";
+            sourceHost = "orm";
+            targetHost = "elucredassa";
+            source = "rpool_orm";
+            target = "rpool_elucredassa/backup/orm";
+          };
+        };
+      };
+
       flake.colmena = {
         meta = {
           nixpkgs = import nixpkgs-unstable {
diff --git a/flake/backup.nix b/flake/backup.nix
new file mode 100644
index 0000000..30bc5bc
--- /dev/null
+++ b/flake/backup.nix
@@ -0,0 +1,126 @@
+{ config, lib, pkgs, ... }:
+let
+  cfg = config.fountain.backup;
+
+  syncOptions = {
+    dataset = lib.mkOption {
+      type = lib.types.str;
+      description = ''
+        The name of the dataset to be synced (not including its parent
+        datasets, if any). This will be the same on the source and target.
+        It must already exist on the source, defined with the
+        {option}`randomcat.services.zfs` module, and not exist on the target.
+      '';
+    };
+    sourceHost = lib.mkOption {
+      type = lib.types.str;
+      description = ''
+        The host from which the dataset should be synced. Must be an entry in
+        {option}`flake.colmena`.
+      '';
+    };
+    targetHost = lib.mkOption {
+      type = lib.types.str;
+      description = ''
+        The host to which the dataset should be synced. Must be an entry in
+        {option}`flake.colmena`.
+      '';
+    };
+    source = lib.mkOption {
+      type = lib.types.str;
+      description = ''
+        The path to the synced dataset in the ZFS namespace on the source host,
+        excluding the component that is the name of the dataset itself.
+      '';
+    };
+    target = lib.mkOption {
+      type = lib.types.str;
+      description = ''
+        The path to the synced dataset in the ZFS namespace on the target host,
+        excluding the component that is the name of the dataset itself. It must
+        already exist, defined with the {option}`randomcat.services.zfs`
+        module.
+      '';
+    };
+  };
+in
+{
+  options.fountain.backup = {
+    keys = lib.mkOption {
+      type = lib.types.attrsOf (lib.types.listOf lib.types.singleLineStr);
+      default = { };
+      description = ''
+        Lists of verbatim OpenSSH public keys that may be used to identify the
+        syncoid user on each target host. The key to each list must be the
+        host's hostname, as listed in {option}`flake.colmena`.
+      '';
+      example = {
+        host = [ "ssh-rsa AAAAB3NzaC1yc2etc/etc/etcjwrsh8e596z6J0l7 example@host" ];
+        bar = [ "ssh-ed25519 AAAAC3NzaCetcetera/etceteraJZMfk3QPfQ foo@bar" ];
+      };
+    };
+
+    sync = lib.mkOption {
+      type = lib.types.attrsOf (lib.types.submodule { options = syncOptions; });
+      default = { };
+      description = ''
+        Details of ZFS datasets whose snapshots should be synced from machine
+        to machine using syncoid. Syncoid will run hourly at 15 past the hour
+        and copy all ZFS snapshots from the source dataset to the target
+        dataset (recursing into child datasets).
+        
+        See descriptions for the individual options for more details. The name
+        of each attribute in this set is arbitrary and used to generate systemd
+        unit names.
+
+        This module does not actually cause snapshots to be taken; sanoid must
+        be configured separately to do this.
+      '';
+      example = {
+        "orm-state" = {
+          dataset = "state";
+          sourceHost = "orm";
+          targetHost = "elucredassa";
+          source = "rpool_orm";
+          target = "rpool_elucredassa/backup/orm";
+        };
+      };
+    };
+  };
+
+  # TODO: add some assertions to verify the options
+
+  config.flake.colmena = lib.mkMerge (lib.mapAttrsToList
+    (name: sync: {
+      ${sync.sourceHost} = { pkgs, ... }: {
+        randomcat.services.zfs.datasets."${sync.source}/${sync.dataset}".zfsPermissions.users.backup = [ "hold" "send" ];
+        users.users.backup = {
+          group = "backup";
+          isSystemUser = true;
+          useDefaultShell = true;
+          openssh.authorizedKeys.keys = cfg.keys.${sync.targetHost};
+          packages = with pkgs; [ mbuffer lzop ]; # syncoid uses these if available but doesn't pull them in automatically
+        };
+        users.groups.backup = { };
+      };
+
+      ${sync.targetHost} = {
+        randomcat.services.zfs.datasets."${sync.target}".zfsPermissions.users.syncoid = [ "mount" "create" "receive" "recordsize" ];
+        services.syncoid = {
+          enable = true;
+          interval = "*-*-* *:15:00";
+          commonArgs = [ "--no-sync-snap" ];
+          commands = {
+            ${name} = {
+              source = "backup@${config.flake.nixosConfigurations.${sync.sourceHost}.config.networking.fqdn}:${sync.source}/${sync.dataset}";
+              target = "${sync.target}/${sync.dataset}";
+              recursive = true;
+              recvOptions = "ux recordsize o compression=lz4";
+            };
+          };
+        };
+      };
+    })
+    cfg.sync
+  );
+}
diff --git a/flake/default.nix b/flake/default.nix
index 65b028d..b567fab 100644
--- a/flake/default.nix
+++ b/flake/default.nix
@@ -1,5 +1,6 @@
 {
   imports = [
+    ./backup.nix
     ./colmena.nix
   ];
 }
diff --git a/hosts/elucredassa/default.nix b/hosts/elucredassa/default.nix
index 147af4d..23c5421 100644
--- a/hosts/elucredassa/default.nix
+++ b/hosts/elucredassa/default.nix
@@ -23,26 +23,10 @@ in
   console.keyMap = "uk";
   services.xserver.xkb.layout = "gb";
 
-  # TODO: modularise this
+  # These are populated by fountain.backup
   randomcat.services.zfs.datasets = {
     "rpool_elucredassa/backup" = { mountpoint = "none"; };
-    "rpool_elucredassa/backup/orm" = {
-      mountpoint = "none";
-      zfsPermissions.users.syncoid = [ "mount" "create" "receive" "recordsize" ];
-    };
-  };
-  services.syncoid = {
-    enable = true;
-    interval = "*-*-* *:15:00";
-    commonArgs = [ "--no-sync-snap" ];
-    commands = {
-      "testing1" = {
-        source = "backup@10.127.1.2:rpool_orm/state";
-        target = "rpool_elucredassa/backup/orm/state";
-        recursive = true;
-        recvOptions = "ux recordsize o compression=lz4";
-      };
-    };
+    "rpool_elucredassa/backup/orm" = { mountpoint = "none"; };
   };
 
   qenya.services.distributed-builds = {
diff --git a/hosts/orm/default.nix b/hosts/orm/default.nix
index c892c20..0db3632 100644
--- a/hosts/orm/default.nix
+++ b/hosts/orm/default.nix
@@ -31,19 +31,6 @@
     useTemplate = [ "production" ];
     recursive = "zfs";
   };
-  
-  # TODO: modularise this
-  randomcat.services.zfs.datasets."rpool_orm/state".zfsPermissions.users.backup = [ "hold" "send" ];
-  users.users.backup = {
-    group = "backup";
-    isSystemUser = true;
-    useDefaultShell = true;
-    openssh.authorizedKeys.keys = [
-      "ssh-ed25519 AAAAC3NzaC1lZDI1NTE5AAAAIOFa3hjej6KGmS2aQ4s46Y7U8pN4yyR2FuMofpHRwXNk syncoid@elucredassa"
-    ];
-    packages = with pkgs; [ mbuffer lzop ]; # syncoid uses these if available but doesn't pull them in automatically
-  };
-  users.groups.backup = { };
 
   qenya.services.actual = {
     enable = true;