feat: patch results on the wire - ApplyUpdatesCmd scope, reboot and deadline, PatchResult, inventory boot time

grpc/pb/patch_test.go

@@ -0,0 +1,46 @@
+package pb
+
+import (
+	"encoding/json"
+	"testing"
+)
+
+// An empty ApplyUpdatesCmd must stay an empty object on the wire, so an old
+// agent and a new server, or a new agent and an old server, agree that it
+// means "install everything, no reboot, no deadline".
+func TestApplyUpdatesCmdEmptyIsEmptyObject(t *testing.T) {
+	b, err := json.Marshal(ApplyUpdatesCmd{})
+	if err != nil {
+		t.Fatal(err)
+	}
+	if string(b) != "{}" {
+		t.Fatalf("got %s, want {}", b)
+	}
+}
+
+func TestPatchResultRoundTrip(t *testing.T) {
+	in := AgentMessage{PatchResult: &PatchResult{
+		CommandId: "c1", Status: PatchStatusOK, OutputTail: "done",
+		PendingAfter: 0, RebootRequired: true, Rebooting: true,
+	}}
+	b, err := json.Marshal(in)
+	if err != nil {
+		t.Fatal(err)
+	}
+	var out AgentMessage
+	if err := json.Unmarshal(b, &out); err != nil {
+		t.Fatal(err)
+	}
+	if out.PatchResult == nil || *out.PatchResult != *in.PatchResult {
+		t.Fatalf("round trip lost data: %+v", out.PatchResult)
+	}
+}
+
+func TestInventoryBootTimeOnWire(t *testing.T) {
+	b, _ := json.Marshal(InventoryReport{BootTimeUnix: 1757800000})
+	var m map[string]any
+	_ = json.Unmarshal(b, &m)
+	if m["boot_time_unix"] != float64(1757800000) {
+		t.Fatalf("boot_time_unix missing: %s", b)
+	}
+}

grpc/pb/vantage.pb.go

@@ -123,6 +123,7 @@ type InventoryReport struct {
 	Partitions     []PartitionReport `json:"partitions,omitempty"`
 	Kernel         string            `json:"kernel,omitempty"`
 	RebootRequired bool              `json:"reboot_required,omitempty"`
+	BootTimeUnix   int64             `json:"boot_time_unix,omitempty"` // every report; proves a reboot happened
 }
 type InventoryReportResponse struct{}
 
@@ -161,7 +162,40 @@ type ReportChecksRequest struct {
 }
 type ReportChecksResponse struct{}
 
-type ApplyUpdatesCmd struct{}
+// ApplyUpdatesCmd installs pending OS updates. The zero value means what the
+// command always meant: every pending update, no reboot, no deadline. That is
+// what keeps old servers and new agents, and new servers and old agents,
+// compatible - but only in that direction for Scope: an agent that predates
+// these fields installs everything even when asked for security only, which
+// is why the control plane gates on agent version before sending a scope.
+type ApplyUpdatesCmd struct {
+	Scope            string `json:"scope,omitempty"`              // "" or PatchScopeAll | PatchScopeSecurity
+	RebootIfRequired bool   `json:"reboot_if_required,omitempty"` // reboot only if the OS reports one is owed
+	DeadlineUnix     int64  `json:"deadline_unix,omitempty"`      // 0 = none; the agent caps the upgrade at 2h
+}
+
+const (
+	PatchScopeAll      = "all"
+	PatchScopeSecurity = "security"
+
+	PatchStatusOK          = "ok"
+	PatchStatusFailed      = "failed"
+	PatchStatusUnsupported = "unsupported"
+	PatchStatusBusy        = "busy"
+)
+
+// PatchResult answers an ApplyUpdatesCmd. Rebooting is sent immediately before
+// the agent restarts the host, so the control plane knows to wait for a
+// post-boot inventory report rather than a second result.
+type PatchResult struct {
+	CommandId      string `json:"command_id"`
+	Status         string `json:"status"`
+	Message        string `json:"message,omitempty"`
+	OutputTail     string `json:"output_tail,omitempty"` // at most 64KB, newest bytes
+	PendingAfter   int32  `json:"pending_after"`         // -1 when the post-apply check failed
+	RebootRequired bool   `json:"reboot_required,omitempty"`
+	Rebooting      bool   `json:"rebooting,omitempty"`
+}
 
 type OpenProxyCmd struct {
 	ProxyId string `json:"proxy_id"`
@@ -240,6 +274,7 @@ type AgentMessage struct {
 	StepOutput *StepOutputChunk `json:"step_output,omitempty"`
 
 	WorkloadLogsResult *WorkloadLogsResult `json:"workload_logs_result,omitempty"`
+	PatchResult        *PatchResult        `json:"patch_result,omitempty"`
 }
 
 type AgentReady struct{}

proto/vantage/v1/vantage.proto

@@ -109,6 +109,7 @@ message AgentMessage {
     StepResult step_result = 5;
     StepOutputChunk step_output = 6;
     WorkloadLogsResult workload_logs_result = 7;
+    PatchResult patch_result = 8;
   }
 }
 
@@ -168,9 +169,9 @@ message InventoryReport {
   uint64 swap_used      = 7;
   repeated PartitionReport partitions = 8;
   string kernel         = 9;
-  // Set on static snapshots only. The agent never reboots; it reports that one
-  // is owed and leaves the decision to a person or a workflow.
+  // Set on static snapshots only. The agent reboots a host only when an ApplyUpdatesCmd asks it to and the OS reports a reboot is owed.
   bool   reboot_required = 10;
+  int64  boot_time_unix = 11;  // every report; proves a reboot happened
 }
 
 message InventoryReportResponse {
@@ -220,7 +221,19 @@ message ReportChecksResponse {
 }
 
 message ApplyUpdatesCmd {
+  string scope = 1;              // "" or "all" | "security"
+  bool   reboot_if_required = 2; // reboot only if the OS reports one is owed
+  int64  deadline_unix = 3;      // 0 = none; the agent caps the upgrade at 2h
+}
 
+message PatchResult {
+  string command_id = 1;
+  string status = 2;             // ok | failed | unsupported | busy
+  string message = 3;
+  string output_tail = 4;        // at most 64KB, newest bytes kept
+  int32  pending_after = 5;      // -1 when the post-apply check failed
+  bool   reboot_required = 6;
+  bool   rebooting = 7;          // sent just before the agent reboots itself
 }
 
 message ServerCommand {