DAOS-16312 control: Always use --force for dmg system stop

Allow-unstable-test: true Features: control Signed-off-by: Tom Nabarro <[email protected]>
daos-stack · Jan 29, 2025 · 6e9e271 · 6e9e271
1 parent e1393d8
commit 6e9e271
Show file tree

Hide file tree

Showing 3 changed files with 19 additions and 7 deletions.
diff --git a/src/control/cmd/dmg/system.go b/src/control/cmd/dmg/system.go
@@ -1,5 +1,6 @@
 //
 // (C) Copyright 2019-2023 Intel Corporation.
+// (C) Copyright 2025 Hewlett Packard Enterprise Development LP
 //
 // SPDX-License-Identifier: BSD-2-Clause-Patent
 //
@@ -185,7 +186,8 @@ func (cmd *systemStopCmd) Execute(_ []string) (errOut error) {
 	if err := cmd.validateHostsRanks(); err != nil {
 		return err
 	}
-	req := &control.SystemStopReq{Force: cmd.Force}
+	// DAOS-16312: Always use force when stopping ranks.
+	req := &control.SystemStopReq{Force: true}
 	req.Hosts.Replace(&cmd.Hosts.HostSet)
 	req.Ranks.Replace(&cmd.Ranks.RankSet)
 

diff --git a/src/control/cmd/dmg/system_test.go b/src/control/cmd/dmg/system_test.go
@@ -1,5 +1,6 @@
 //
 // (C) Copyright 2019-2023 Intel Corporation.
+// (C) Copyright 2025 Hewlett Packard Enterprise Development LP
 //
 // SPDX-License-Identifier: BSD-2-Clause-Patent
 //
@@ -144,7 +145,7 @@ func TestDmg_SystemCommands(t *testing.T) {
 			"system stop with no arguments",
 			"system stop",
 			strings.Join([]string{
-				printRequest(t, &control.SystemStopReq{}),
+				printRequest(t, &control.SystemStopReq{Force: true}),
 			}, " "),
 			nil,
 		},
@@ -160,23 +161,26 @@ func TestDmg_SystemCommands(t *testing.T) {
 			"system stop with single rank",
 			"system stop --ranks 0",
 			strings.Join([]string{
-				printRequest(t, withRanks(&control.SystemStopReq{}, 0)),
+				printRequest(t, withRanks(&control.SystemStopReq{Force: true}, 0)),
 			}, " "),
 			nil,
 		},
 		{
 			"system stop with multiple ranks",
 			"system stop --ranks 0,1,4",
 			strings.Join([]string{
-				printRequest(t, withRanks(&control.SystemStopReq{}, 0, 1, 4)),
+				printRequest(t,
+					withRanks(&control.SystemStopReq{Force: true}, 0, 1, 4)),
 			}, " "),
 			nil,
 		},
 		{
 			"system stop with multiple hosts",
 			"system stop --rank-hosts bar9,foo-[0-100]",
 			strings.Join([]string{
-				printRequest(t, withHosts(&control.SystemStopReq{}, "foo-[0-100]", "bar9")),
+				printRequest(t,
+					withHosts(&control.SystemStopReq{Force: true},
+						"foo-[0-100]", "bar9")),
 			}, " "),
 			nil,
 		},

diff --git a/src/control/server/mgmt_system.go b/src/control/server/mgmt_system.go
@@ -831,9 +831,12 @@ func (svc *mgmtSvc) SystemStop(ctx context.Context, req *mgmtpb.SystemStopReq) (
 	if err != nil {
 		return nil, err
 	}
+	svc.log.Debugf("system-stop fanout request: %+v", fReq)
 
-	// First phase: Prepare the ranks for shutdown, but only if the request
-	// is for an unforced full system stop.
+	// First phase: Prepare the ranks for shutdown, but only if the request is for an unforced
+	// full system stop.
+	// DAOS-16312: Note that if force is always specified in request, PrepShutdown will never
+	// be called.
 	if fReq.FullSystem && !fReq.Force {
 		fReq.Method = control.PrepShutdownRanks
 		fResp, _, err = svc.rpcFanout(ctx, fReq, fResp, true)
@@ -942,6 +945,7 @@ func (svc *mgmtSvc) SystemStart(ctx context.Context, req *mgmtpb.SystemStartReq)
 	if err != nil {
 		return nil, err
 	}
+	svc.log.Debugf("system-start fanout request: %+v", fReq)
 
 	fReq.CheckMode = req.CheckMode
 	fReq.Method = control.StartRanks
@@ -963,6 +967,7 @@ func (svc *mgmtSvc) SystemExclude(ctx context.Context, req *mgmtpb.SystemExclude
 	if err := svc.checkLeaderRequest(wrapCheckerReq(req)); err != nil {
 		return nil, err
 	}
+	svc.log.Debug("Received SystemExclude RPC")
 
 	if req.Hosts == "" && req.Ranks == "" {
 		return nil, errors.New("no hosts or ranks specified")
@@ -972,6 +977,7 @@ func (svc *mgmtSvc) SystemExclude(ctx context.Context, req *mgmtpb.SystemExclude
 	if err != nil {
 		return nil, err
 	}
+	svc.log.Debugf("system-exclude fanout request: %+v", fReq)
 
 	if fResp.AbsentHosts.Count() > 0 {
 		return nil, errors.Errorf("invalid host(s): %s", fResp.AbsentHosts.String())