nixos/ceph: run unprivileged, use StateDirectory and tmpfiles, don't pass extraServiceConfig

Don't pass user and group to ceph, and rely on it to drop ceps, but let
systemd handle running it as the appropriate user.

This also inlines the extraServiceConfig into the makeService function,
as we have conditionals depending on daemonType there anyways.

Use StateDirectory to create directories in
/var/lib/ceph/${daemonType}/${clusterName}-${daemonId}.

There previously was a condition on daemonType being one of mds,mon,rgw
or mgr. We only instantiate makeServices with these types, and "osd" was
special.
In the osd case, test examples suggest it'd be in something like
/var/lib/ceph/osd/ceph-${cfg.osd0.name} - so it's not special at all,
but exactly like the pattern for the others.

During initialization, we also need these folders, before the unit is
started up. Move the mkdir -p commands in the vm tests to the line
immediately before they're required.

3 files changed, 36 insertions, 49 deletions

diff --git a/nixos/modules/services/network-filesystems/ceph.nix b/nixos/modules/services/network-filesystems/ceph.nix
index 887a6ff925d..ef0d2d90a1a 100644
--- a/nixos/modules/services/network-filesystems/ceph.nix
+++ b/nixos/modules/services/network-filesystems/ceph.nix
@@ -9,12 +9,12 @@ let
   expandCamelCase = replaceStrings upperChars (map (s: " ${s}") lowerChars);
   expandCamelCaseAttrs = mapAttrs' (name: value: nameValuePair (expandCamelCase name) value);
 
-  makeServices = (daemonType: daemonIds: extraServiceConfig:
+  makeServices = (daemonType: daemonIds:
     mkMerge (map (daemonId:
-      { "ceph-${daemonType}-${daemonId}" = makeService daemonType daemonId cfg.global.clusterName pkgs.ceph extraServiceConfig; })
+      { "ceph-${daemonType}-${daemonId}" = makeService daemonType daemonId cfg.global.clusterName pkgs.ceph; })
       daemonIds));
 
-  makeService = (daemonType: daemonId: clusterName: ceph: extraServiceConfig: {
+  makeService = (daemonType: daemonId: clusterName: ceph: {
     enable = true;
     description = "Ceph ${builtins.replaceStrings lowerChars upperChars daemonType} daemon ${daemonId}";
     after = [ "network-online.target" "time-sync.target" ] ++ optional (daemonType == "osd") "ceph-mon.target";
@@ -22,6 +22,8 @@ let
     partOf = [ "ceph-${daemonType}.target" ];
     wantedBy = [ "ceph-${daemonType}.target" ];
 
+    path = [ pkgs.getopt ];
+
     serviceConfig = {
       LimitNOFILE = 1048576;
       LimitNPROC = 1048576;
@@ -34,22 +36,22 @@ let
       Restart = "on-failure";
       StartLimitBurst = "5";
       StartLimitInterval = "30min";
+      StateDirectory = "ceph/${if daemonType == "rgw" then "radosgw" else daemonType}/${clusterName}-${daemonId}";
+      User = "ceph";
+      Group = if daemonType == "osd" then "disk" else "ceph";
       ExecStart = ''${ceph.out}/bin/${if daemonType == "rgw" then "radosgw" else "ceph-${daemonType}"} \
-                    -f --cluster ${clusterName} --id ${daemonId} --setuser ceph \
-                    --setgroup ${if daemonType == "osd" then "disk" else "ceph"}'';
-    } // extraServiceConfig
-      // optionalAttrs (daemonType == "osd") { ExecStartPre = ''${ceph.lib}/libexec/ceph/ceph-osd-prestart.sh \
-                                                              --id ${daemonId} --cluster ${clusterName}''; };
-    } // optionalAttrs (builtins.elem daemonType [ "mds" "mon" "rgw" "mgr" ]) {
-      preStart = ''
-        daemonPath="/var/lib/ceph/${if daemonType == "rgw" then "radosgw" else daemonType}/${clusterName}-${daemonId}"
-        if [ ! -d $daemonPath ]; then
-          mkdir -m 755 -p $daemonPath
-          chown -R ceph:ceph $daemonPath
-        fi
-      '';
-    } // optionalAttrs (daemonType == "osd") { path = [ pkgs.getopt ]; }
-  );
+                    -f --cluster ${clusterName} --id ${daemonId}'';
+    } // optionalAttrs (daemonType == "osd") {
+      ExecStartPre = ''${ceph.lib}/libexec/ceph/ceph-osd-prestart.sh --id ${daemonId} --cluster ${clusterName}'';
+      StartLimitBurst = "30";
+      RestartSec = "20s";
+      PrivateDevices = "no"; # osd needs disk access
+    } // optionalAttrs ( daemonType == "mon") {
+      RestartSec = "10";
+    } // optionalAttrs (lib.elem daemonType ["mgr" "mds"]) {
+      StartLimitBurst = "3";
+    };
+  });
 
   makeTarget = (daemonType:
     {
@@ -377,14 +379,11 @@ in
 
     systemd.services = let
       services = []
-        ++ optional cfg.mon.enable (makeServices "mon" cfg.mon.daemons { RestartSec = "10"; })
-        ++ optional cfg.mds.enable (makeServices "mds" cfg.mds.daemons { StartLimitBurst = "3"; })
-        ++ optional cfg.osd.enable (makeServices "osd" cfg.osd.daemons { StartLimitBurst = "30";
-                                                                         RestartSec = "20s";
-                                                                         PrivateDevices = "no"; # osd needs disk access
-                                                                       })
-        ++ optional cfg.rgw.enable (makeServices "rgw" cfg.rgw.daemons { })
-        ++ optional cfg.mgr.enable (makeServices "mgr" cfg.mgr.daemons { StartLimitBurst = "3"; });
+        ++ optional cfg.mon.enable (makeServices "mon" cfg.mon.daemons)
+        ++ optional cfg.mds.enable (makeServices "mds" cfg.mds.daemons)
+        ++ optional cfg.osd.enable (makeServices "osd" cfg.osd.daemons)
+        ++ optional cfg.rgw.enable (makeServices "rgw" cfg.rgw.daemons)
+        ++ optional cfg.mgr.enable (makeServices "mgr" cfg.mgr.daemons);
       in
         mkMerge services;
 
@@ -403,7 +402,9 @@ in
     systemd.tmpfiles.rules = [
       "d /etc/ceph - ceph ceph - -"
       "d /run/ceph 0770 ceph ceph -"
-      "d /var/lib/ceph - ceph ceph - -"
-    ];
+      "d /var/lib/ceph - ceph ceph - -"]
+    ++ optionals cfg.mgr.enable [ "d /var/lib/ceph/mgr - ceph ceph - -"]
+    ++ optionals cfg.mon.enable [ "d /var/lib/ceph/mon - ceph ceph - -"]
+    ++ optionals cfg.osd.enable [ "d /var/lib/ceph/osd - ceph ceph - -"];
   };
 }
diff --git a/nixos/tests/ceph-multi-node.nix b/nixos/tests/ceph-multi-node.nix
index 13d0851366e..c34b134de53 100644
--- a/nixos/tests/ceph-multi-node.nix
+++ b/nixos/tests/ceph-multi-node.nix
@@ -114,18 +114,6 @@ let
     $osd0->waitForUnit("network.target");
     $osd1->waitForUnit("network.target");
 
-    # Create the ceph-related directories
-    $monA->mustSucceed(
-      "mkdir -p /var/lib/ceph/mgr/ceph-${cfg.monA.name}",
-      "mkdir -p /var/lib/ceph/mon/ceph-${cfg.monA.name}",
-    );
-    $osd0->mustSucceed(
-      "mkdir -p /var/lib/ceph/osd/ceph-${cfg.osd0.name}",
-    );
-    $osd1->mustSucceed(
-      "mkdir -p /var/lib/ceph/osd/ceph-${cfg.osd1.name}",
-    );
-
     # Bootstrap ceph-mon daemon
     $monA->mustSucceed(
       "sudo -u ceph ceph-authtool --create-keyring /tmp/ceph.mon.keyring --gen-key -n mon. --cap mon 'allow *'",
@@ -133,6 +121,7 @@ let
       "sudo -u ceph ceph-authtool /tmp/ceph.mon.keyring --import-keyring /etc/ceph/ceph.client.admin.keyring",
       "monmaptool --create --add ${cfg.monA.name} ${cfg.monA.ip} --fsid ${cfg.clusterId} /tmp/monmap",
       "sudo -u ceph ceph-mon --mkfs -i ${cfg.monA.name} --monmap /tmp/monmap --keyring /tmp/ceph.mon.keyring",
+      "sudo -u ceph mkdir -p /var/lib/ceph/mgr/ceph-${cfg.monA.name}/",
       "sudo -u ceph touch /var/lib/ceph/mon/ceph-${cfg.monA.name}/done",
       "systemctl start ceph-mon-${cfg.monA.name}"
     );
@@ -159,12 +148,14 @@ let
     # Bootstrap both OSDs
     $osd0->mustSucceed(
       "mkfs.xfs /dev/vdb",
+      "mkdir -p /var/lib/ceph/osd/ceph-${cfg.osd0.name}",
       "mount /dev/vdb /var/lib/ceph/osd/ceph-${cfg.osd0.name}",
       "ceph-authtool --create-keyring /var/lib/ceph/osd/ceph-${cfg.osd0.name}/keyring --name osd.${cfg.osd0.name} --add-key ${cfg.osd0.key}",
       "echo '{\"cephx_secret\": \"${cfg.osd0.key}\"}' | ceph osd new ${cfg.osd0.uuid} -i -",
     );
     $osd1->mustSucceed(
       "mkfs.xfs /dev/vdb",
+      "mkdir -p /var/lib/ceph/osd/ceph-${cfg.osd1.name}",
       "mount /dev/vdb /var/lib/ceph/osd/ceph-${cfg.osd1.name}",
       "ceph-authtool --create-keyring /var/lib/ceph/osd/ceph-${cfg.osd1.name}/keyring --name osd.${cfg.osd1.name} --add-key ${cfg.osd1.key}",
       "echo '{\"cephx_secret\": \"${cfg.osd1.key}\"}' | ceph osd new ${cfg.osd1.uuid} -i -"
diff --git a/nixos/tests/ceph-single-node.nix b/nixos/tests/ceph-single-node.nix
index cb37b4b43bc..5938300ebb4 100644
--- a/nixos/tests/ceph-single-node.nix
+++ b/nixos/tests/ceph-single-node.nix
@@ -77,14 +77,6 @@ let
 
     $monA->waitForUnit("network.target");
 
-    # Create the ceph-related directories
-    $monA->mustSucceed(
-      "mkdir -p /var/lib/ceph/mgr/ceph-${cfg.monA.name}",
-      "mkdir -p /var/lib/ceph/mon/ceph-${cfg.monA.name}",
-      "mkdir -p /var/lib/ceph/osd/ceph-${cfg.osd0.name}",
-      "mkdir -p /var/lib/ceph/osd/ceph-${cfg.osd1.name}",
-    );
-
     # Bootstrap ceph-mon daemon
     $monA->mustSucceed(
       "sudo -u ceph ceph-authtool --create-keyring /tmp/ceph.mon.keyring --gen-key -n mon. --cap mon 'allow *'",
@@ -101,8 +93,9 @@ let
     # Can't check ceph status until a mon is up
     $monA->succeed("ceph -s | grep 'mon: 1 daemons'");
 
-    # Start the ceph-mgr daemon, it has no deps and hardly any setup
+    # Start the ceph-mgr daemon, after copying in the keyring
     $monA->mustSucceed(
+      "sudo -u ceph mkdir -p /var/lib/ceph/mgr/ceph-${cfg.monA.name}/",
       "ceph auth get-or-create mgr.${cfg.monA.name} mon 'allow profile mgr' osd 'allow *' mds 'allow *' > /var/lib/ceph/mgr/ceph-${cfg.monA.name}/keyring",
       "systemctl start ceph-mgr-${cfg.monA.name}"
     );
@@ -114,7 +107,9 @@ let
     $monA->mustSucceed(
       "mkfs.xfs /dev/vdb",
       "mkfs.xfs /dev/vdc",
+      "mkdir -p /var/lib/ceph/osd/ceph-${cfg.osd0.name}",
       "mount /dev/vdb /var/lib/ceph/osd/ceph-${cfg.osd0.name}",
+      "mkdir -p /var/lib/ceph/osd/ceph-${cfg.osd1.name}",
       "mount /dev/vdc /var/lib/ceph/osd/ceph-${cfg.osd1.name}",
       "ceph-authtool --create-keyring /var/lib/ceph/osd/ceph-${cfg.osd0.name}/keyring --name osd.${cfg.osd0.name} --add-key ${cfg.osd0.key}",
       "ceph-authtool --create-keyring /var/lib/ceph/osd/ceph-${cfg.osd1.name}/keyring --name osd.${cfg.osd1.name} --add-key ${cfg.osd1.key}",