Merge pull request #2699 from kc284/master

CA-215578: Fixed issue where halted control domains were deleted from…

ocaml/test/OMakefile

@@ -42,6 +42,7 @@ OCAML_OBJS = \
 	test_map_check \
 	test_pool_apply_edition \
 	test_pool_license \
+	test_pool_restore_database \
 	test_platformdata \
 	test_sm_features \
 	test_gpu_group \

ocaml/test/suite.ml

@@ -34,6 +34,7 @@ let base_suite =
 			Test_map_check.test;
 			Test_pool_apply_edition.test;
 			Test_pool_license.test;
+			Test_pool_restore_database.test;
 			Test_platformdata.test;
 			Test_sm_features.test;
 			Test_gpu_group.test;

ocaml/test/test_pool_restore_database.ml

@@ -0,0 +1,48 @@
+(*
+ * Copyright (C) 2016 Citrix Systems Inc.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU Lesser General Public License as published
+ * by the Free Software Foundation; version 2.1 only. with the special
+ * exception on linking described in file LICENSE.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU Lesser General Public License for more details.
+ *)
+
+open OUnit
+open Test_common
+open Dbsync_master
+
+let run_test vm =
+  let (nameLabel, isControlDomain, powerState) = vm in
+  let __context = Test_common.make_test_database () in
+  let vm = Test_common.make_vm ~__context ~name_label:nameLabel ()
+  in
+  Db.VM.set_is_control_domain ~__context ~self:vm ~value:isControlDomain;
+  Db.VM.set_power_state ~__context ~self:vm ~value:powerState;
+  Db.VM.set_resident_on ~__context ~self:vm ~value:Ref.null;
+
+  Dbsync_master.reset_vms_running_on_missing_hosts ~__context;
+
+  assert_equal ~msg:(Printf.sprintf "The VM %s is not halted" nameLabel)
+    (Db.VM.get_power_state ~__context ~self:vm) `Halted
+
+let test_cases = [
+  ("vm1", false, `Running);
+  ("vm2", false, `Halted);
+  ("dom1", true, `Running);
+  ("dom2", true, `Halted)
+]
+
+let test_reset_vms_on_missing_host () =
+  List.iter run_test test_cases
+
+
+let test =
+  "pool_restore_database" >:::
+    [
+      "test_reset_vms_on_missing_host" >:: test_reset_vms_on_missing_host;
+    ]

ocaml/xapi/dbsync_master.ml

@@ -81,18 +81,13 @@ let refresh_console_urls ~__context =
     If these are control domains we destroy them, otherwise we reset them to Halted. *)
 let reset_vms_running_on_missing_hosts ~__context =
   List.iter (fun vm ->
-	       let vm_r = Db.VM.get_record ~__context ~self:vm in
-	       let valid_resident_on = Db.is_valid_ref __context vm_r.API.vM_resident_on in
-	       if not valid_resident_on then begin
-		 if vm_r.API.vM_is_control_domain then begin
-		   info "Deleting control domain VM uuid '%s' ecause VM.resident_on refers to a Host which is nolonger in the Pool" vm_r.API.vM_uuid;
-		   Db.VM.destroy ~__context ~self:vm
-		 end else if vm_r.API.vM_power_state = `Running then begin
-		   let msg = Printf.sprintf "Resetting VM uuid '%s' to Halted because VM.resident_on refers to a Host which is nolonger in the Pool" vm_r.API.vM_uuid in
-		   info "%s" msg;
-		   Helpers.log_exn_continue msg (fun () -> Xapi_vm_lifecycle.force_state_reset ~__context ~self:vm ~value:`Halted) ()
-		 end
-	       end) (Db.VM.get_all ~__context)
+    let vm_r = Db.VM.get_record ~__context ~self:vm in
+    let valid_resident_on = Db.is_valid_ref __context vm_r.API.vM_resident_on in
+    if (not valid_resident_on) && (vm_r.API.vM_power_state = `Running) then begin
+      let msg = Printf.sprintf "Resetting VM uuid '%s' to Halted because VM.resident_on refers to a Host which is no longer in the Pool" vm_r.API.vM_uuid in
+      info "%s" msg;
+      Helpers.log_exn_continue msg (fun () -> Xapi_vm_lifecycle.force_state_reset ~__context ~self:vm ~value:`Halted) ()
+    end) (Db.VM.get_all ~__context)
 
 (** Release 'locks' on VMs in the Halted state: ie {VBD,VIF}.{currently_attached,reserved}
     Note that the {allowed,current}_operations fields are non-persistent so blanked on *master* startup (not slave)

ocaml/xapi/xapi_vm_lifecycle.ml

@@ -497,7 +497,7 @@ let force_state_reset_keep_current_operations ~__context ~self ~value:state =
 	if state = `Halted then begin
 		(* archive the rrd for this vm *)
 		let vm_uuid = Db.VM.get_uuid ~__context ~self in
-		Rrdd.archive_rrd ~vm_uuid ~remote_address:(try Some (Pool_role.get_master_address ()) with _ -> None)
+		log_and_ignore_exn (fun () -> Rrdd.archive_rrd ~vm_uuid ~remote_address:(try Some (Pool_role.get_master_address ()) with _ -> None))
 	end;
 
 	Db.VM.set_power_state ~__context ~self ~value:state;