| 1 | package hostbox |
| 2 | |
| 3 | import ( |
| 4 | "context" |
| 5 | "crypto/rand" |
| 6 | "crypto/sha256" |
| 7 | "database/sql" |
| 8 | "encoding/hex" |
| 9 | "encoding/json" |
| 10 | "errors" |
| 11 | "fmt" |
| 12 | "os" |
| 13 | "path/filepath" |
| 14 | "reflect" |
| 15 | "strings" |
| 16 | |
| 17 | "github.com/carlosframework/platform/internal/hostclaim" |
| 18 | "github.com/carlosframework/platform/internal/hostguard" |
| 19 | "github.com/carlosframework/platform/internal/hostwrite" |
| 20 | "github.com/carlosframework/platform/internal/provision" |
| 21 | "github.com/carlosframework/platform/internal/registry" |
| 22 | ) |
| 23 | |
| 24 | type ReconcileAdapter struct { |
| 25 | RequireTimerInventory bool |
| 26 | DataRoot string |
| 27 | RemoteRetire func(context.Context, hostguard.FenceTarget) error |
| 28 | Census *EmptyInstallationCensus |
| 29 | Reconciler Reconciler |
| 30 | Registry *registry.Registry |
| 31 | Inventory *StopInventoryDB |
| 32 | Box, Enrollment, DirectoryRevision string |
| 33 | Stop func(context.Context, registry.Route) (hostguard.FenceTarget, error) |
| 34 | Retirement TypedRetirement |
| 35 | } |
| 36 | |
| 37 | func (a ReconcileAdapter) Configure(cfg *provision.ReconcileConfig) { |
| 38 | cfg.ClaimEnabled = a.Reconciler.Writes.Enabled |
| 39 | cfg.ClaimInstall = a.Install |
| 40 | cfg.ClaimRetire = a.Retire |
| 41 | } |
| 42 | func (a ReconcileAdapter) Install(ctx context.Context, rec provision.ClaimRecord, write func(registry.ClaimOwnership) error) error { |
| 43 | if a.Registry == nil || a.Inventory == nil || a.Reconciler.Claims == nil || write == nil || rec.Value == nil || a.Box == "" || a.Enrollment == "" || a.DirectoryRevision == "" { |
| 44 | return ErrUnsupported |
| 45 | } |
| 46 | body, etag, err := a.Reconciler.Claims.Get(ctx, rec.Key) |
| 47 | if err != nil { |
| 48 | return err |
| 49 | } |
| 50 | current := reflect.New(reflect.TypeOf(rec.Value)) |
| 51 | if err = json.Unmarshal(body, current.Interface()); err != nil { |
| 52 | return err |
| 53 | } |
| 54 | if !reflect.DeepEqual(current.Elem().Interface(), rec.Value) { |
| 55 | return ErrChanged |
| 56 | } |
| 57 | inc, err := hostclaim.Incarnation(rec.Account, rec.App, rec.Host, rec.CreatedAt) |
| 58 | if err != nil { |
| 59 | return err |
| 60 | } |
| 61 | // The claim inventory records where this box's files are: the route's own |
| 62 | // paths say so (sidecars are always flat). |
| 63 | layout := "flat" |
| 64 | if rec.Family != "sidecar" && provision.NestedPaths(a.Reconciler.Guard.DataDir, rec.Route) { |
| 65 | layout = "nested" |
| 66 | } |
| 67 | selected := RecordSelection{Host: rec.Host, Key: rec.Key, ETag: etag, Incarnation: inc, Family: rec.Family, Layout: layout, DryRun: rec.DryRun} |
| 68 | reconciler := a.Reconciler |
| 69 | priorCheck := reconciler.CheckRecord |
| 70 | reconciler.CheckRecord = func(ctx context.Context, selected RecordSelection) error { |
| 71 | _, currentETag, err := reconciler.Claims.Get(ctx, selected.Key) |
| 72 | if err != nil { |
| 73 | return err |
| 74 | } |
| 75 | if currentETag != selected.ETag { |
| 76 | return ErrChanged |
| 77 | } |
| 78 | if priorCheck != nil { |
| 79 | return priorCheck(ctx, selected) |
| 80 | } |
| 81 | return nil |
| 82 | } |
| 83 | if reconciler.CheckHeld == nil && a.Census != nil { |
| 84 | // The hold-only path's local check is the census's own (Task 10b). |
| 85 | reconciler.CheckHeld = a.Census.CheckHeld |
| 86 | } |
| 87 | if a.Inventory.remote != nil && rec.Route.Status == "landing" { |
| 88 | if rec.DryRun { |
| 89 | return nil |
| 90 | } |
| 91 | obs, err := hostclaim.Read(ctx, a.Reconciler.Claims, rec.Host) |
| 92 | if err != nil { |
| 93 | return err |
| 94 | } |
| 95 | return a.RegisterLanding(ctx, obs) |
| 96 | } |
| 97 | if a.Inventory.remote != nil { |
| 98 | return a.installRemote(ctx, rec, selected, reconciler, write) |
| 99 | } |
| 100 | return reconciler.Install(ctx, selected, func() error { |
| 101 | captured, err := a.Reconciler.Guard.Capture(rec.Host) |
| 102 | if err != nil { |
| 103 | return err |
| 104 | } |
| 105 | if captured.Incarnation != inc { |
| 106 | return ErrChanged |
| 107 | } |
| 108 | return a.ownedRow(ctx, rec, inc, selected, captured, write) |
| 109 | }) |
| 110 | } |
| 111 | |
| 112 | // ownedRow is the typed-inventory half of a row write under the host guard |
| 113 | // at captured (extracted unchanged from Install): the installation's |
| 114 | // inventory (initialised by the census when none exists), one route entry |
| 115 | // owned by captured's digest, write(owner), and the entry materialized. |
| 116 | // Install calls it inside Reconciler.Install's selected work; MintHeld |
| 117 | // inside a relocation install's typed work (region HA rulings 44, 46). |
| 118 | func (a ReconcileAdapter) ownedRow(ctx context.Context, rec provision.ClaimRecord, inc string, selected RecordSelection, captured hostguard.Captured, write func(registry.ClaimOwnership) error) error { |
| 119 | in, err := a.Inventory.ReadInstallation(ctx, rec.Host, inc, captured.Stamp) |
| 120 | if errors.Is(err, sql.ErrNoRows) { |
| 121 | if a.Census == nil { |
| 122 | return ErrTypedInventory |
| 123 | } |
| 124 | observed, e := hostclaim.Read(ctx, a.Reconciler.Claims, rec.Host) |
| 125 | if e != nil { |
| 126 | return e |
| 127 | } |
| 128 | if observed.Claim == nil || observed.Claim.Holder == nil || observed.Claim.Holder.Incarnation != captured.Incarnation || observed.Claim.Holder.Stamp != captured.Stamp { |
| 129 | return ErrChanged |
| 130 | } |
| 131 | if e = a.Census.Initialize(ctx, observed, selected); e != nil { |
| 132 | return e |
| 133 | } |
| 134 | in, err = a.Inventory.ReadInstallation(ctx, rec.Host, inc, captured.Stamp) |
| 135 | } |
| 136 | if err != nil { |
| 137 | return err |
| 138 | } |
| 139 | if !in.Complete || in.Box != a.Box || in.Enrollment != a.Enrollment || in.DirectoryRevision != a.DirectoryRevision { |
| 140 | return ErrTypedInventory |
| 141 | } |
| 142 | ownerDigest := sha256.Sum256([]byte(fmt.Sprintf("%s\x00%s\x00%d\x00%s", rec.Host, inc, captured.Stamp, a.Box))) |
| 143 | owner := registry.ClaimOwnership{ClaimIncarnation: inc, ClaimStamp: captured.Stamp, ClaimOwner: hex.EncodeToString(ownerDigest[:])} |
| 144 | entry := -1 |
| 145 | for i, e := range in.Entries { |
| 146 | if e.Kind == "route" && e.RelativePath == rec.Host { |
| 147 | if entry != -1 || e.Ownership != owner.ClaimOwner || e.RowIdentity == "" { |
| 148 | return ErrTypedInventory |
| 149 | } |
| 150 | entry = i |
| 151 | owner.ClaimRowID = e.RowIdentity |
| 152 | } |
| 153 | } |
| 154 | if entry < 0 { |
| 155 | var id [16]byte |
| 156 | if _, err = rand.Read(id[:]); err != nil { |
| 157 | return err |
| 158 | } |
| 159 | owner.ClaimRowID = hex.EncodeToString(id[:]) |
| 160 | entry = len(in.Entries) |
| 161 | in.Entries = append(in.Entries, StopInventoryEntry{Kind: "route", ArtifactID: owner.ClaimRowID, ManagedRoot: "registry", RelativePath: rec.Host, RowIdentity: owner.ClaimRowID, Ownership: owner.ClaimOwner, Materialization: "intended"}) |
| 162 | if err = a.Inventory.WriteStopInventory(ctx, in); err != nil { |
| 163 | return err |
| 164 | } |
| 165 | } |
| 166 | row, err := a.Registry.Get(rec.Host) |
| 167 | if err != nil { |
| 168 | return err |
| 169 | } |
| 170 | if row == nil && in.Entries[entry].Materialization == "materialized" { |
| 171 | return ErrTypedInventory |
| 172 | } |
| 173 | if row != nil && row.ClaimOwnership != owner { |
| 174 | return registry.ErrClaimOwnership |
| 175 | } |
| 176 | if err = write(owner); err != nil { |
| 177 | return err |
| 178 | } |
| 179 | row, err = a.Registry.Get(rec.Host) |
| 180 | if err != nil { |
| 181 | return err |
| 182 | } |
| 183 | if row == nil || row.ClaimOwnership != owner { |
| 184 | return registry.ErrClaimOwnership |
| 185 | } |
| 186 | in.Entries[entry].Materialization = "materialized" |
| 187 | return a.Inventory.WriteStopInventory(ctx, in) |
| 188 | } |
| 189 | |
| 190 | // MintHeld mints, or keeps, the host's row held by op with status under the |
| 191 | // installation captured, for a relocation target's local install (region |
| 192 | // HA rulings 31, 44): the route reconcile would mint for rec, nested |
| 193 | // (ruling 45), with its ClaimOwnership, status and hold_op set in one |
| 194 | // Upsert; an existing row must already be this installation's and held by |
| 195 | // op. It runs inside the install's typed work, first, before any file, |
| 196 | // timer or env is written, so the census initialises the installation |
| 197 | // from an empty box. |
| 198 | func (a ReconcileAdapter) MintHeld(ctx context.Context, rec provision.ClaimRecord, captured hostguard.Captured, status, op string) error { |
| 199 | if !registry.IsHold(status) || op == "" || a.Registry == nil || a.Inventory == nil || a.Box == "" || a.Enrollment == "" || a.DirectoryRevision == "" { |
| 200 | return ErrUnsupported |
| 201 | } |
| 202 | inc, err := hostclaim.Incarnation(rec.Account, rec.App, rec.Host, rec.CreatedAt) |
| 203 | if err != nil { |
| 204 | return err |
| 205 | } |
| 206 | if captured.Incarnation != inc || rec.Family != "instance" || !provision.NestedPaths(a.Reconciler.Guard.DataDir, rec.Route) { |
| 207 | return ErrChanged |
| 208 | } |
| 209 | selected := RecordSelection{Host: rec.Host, Key: rec.Key, Incarnation: inc, Family: rec.Family, Layout: "nested"} |
| 210 | return a.ownedRow(ctx, rec, inc, selected, captured, func(owner registry.ClaimOwnership) error { |
| 211 | row, err := a.Registry.Get(rec.Host) |
| 212 | if err != nil { |
| 213 | return err |
| 214 | } |
| 215 | if row != nil { |
| 216 | if row.ClaimOwnership != owner || row.HoldOp != op || !registry.IsHold(row.Status) { |
| 217 | return registry.ErrClaimOwnership |
| 218 | } |
| 219 | return nil |
| 220 | } |
| 221 | rt := rec.Route |
| 222 | rt.ClaimOwnership, rt.Status, rt.HoldOp = owner, status, op |
| 223 | return a.Registry.Upsert(rt) |
| 224 | }) |
| 225 | } |
| 226 | |
| 227 | func (a ReconcileAdapter) MintHeldUnder(ctx context.Context, rec provision.ClaimRecord, captured hostguard.Captured, status, op string, work hostwrite.Work) error { |
| 228 | if work == nil || a.Inventory == nil { |
| 229 | return ErrUnsupported |
| 230 | } |
| 231 | if a.Inventory.remote == nil { |
| 232 | return work.With(ctx, hostwrite.Identity{}, func() error { return a.MintHeld(ctx, rec, captured, status, op) }) |
| 233 | } |
| 234 | if !registry.IsHold(status) || op == "" || a.Registry == nil || a.Reconciler.Claims == nil || a.Box == "" || a.Enrollment == "" || a.DirectoryRevision == "" || rec.Value == nil { |
| 235 | return ErrUnsupported |
| 236 | } |
| 237 | inc, err := hostclaim.Incarnation(rec.Account, rec.App, rec.Host, rec.CreatedAt) |
| 238 | if err != nil { |
| 239 | return err |
| 240 | } |
| 241 | if captured.Incarnation != inc || rec.Family != "instance" || !provision.NestedPaths(a.Reconciler.Guard.DataDir, rec.Route) { |
| 242 | return ErrChanged |
| 243 | } |
| 244 | body, etag, err := a.Reconciler.Claims.Get(ctx, rec.Key) |
| 245 | if err != nil { |
| 246 | return err |
| 247 | } |
| 248 | current := reflect.New(reflect.TypeOf(rec.Value)) |
| 249 | if err = json.Unmarshal(body, current.Interface()); err != nil { |
| 250 | return err |
| 251 | } |
| 252 | if !reflect.DeepEqual(current.Elem().Interface(), rec.Value) { |
| 253 | return ErrChanged |
| 254 | } |
| 255 | selected := RecordSelection{Host: rec.Host, Key: rec.Key, ETag: etag, Incarnation: inc, Family: rec.Family, Layout: "nested"} |
| 256 | if err := a.prepareRemoteInventory(ctx, selected, captured); err != nil { |
| 257 | return err |
| 258 | } |
| 259 | return a.registerRemote(ctx, selected, captured, work, func(owner registry.ClaimOwnership) error { |
| 260 | row, err := a.Registry.Get(rec.Host) |
| 261 | if err != nil { |
| 262 | return err |
| 263 | } |
| 264 | if row != nil { |
| 265 | if row.ClaimOwnership != owner || row.HoldOp != op || !registry.IsHold(row.Status) { |
| 266 | return registry.ErrClaimOwnership |
| 267 | } |
| 268 | return nil |
| 269 | } |
| 270 | rt := rec.Route |
| 271 | rt.ClaimOwnership, rt.Status, rt.HoldOp = owner, status, op |
| 272 | return a.Registry.Upsert(rt) |
| 273 | }) |
| 274 | } |
| 275 | |
| 276 | func (a ReconcileAdapter) Retire(ctx context.Context, row registry.Route) error { |
| 277 | if a.Registry == nil || a.Inventory == nil || a.Stop == nil || !row.ClaimOwnership.Valid() { |
| 278 | return ErrUnsupported |
| 279 | } |
| 280 | target, err := a.Stop(ctx, row) |
| 281 | if err != nil { |
| 282 | return err |
| 283 | } |
| 284 | if target.Host != row.Host || target.Incarnation != row.ClaimIncarnation || target.Stamp != row.ClaimStamp || target.Box != a.Box { |
| 285 | return ErrChanged |
| 286 | } |
| 287 | |
| 288 | if a.RemoteRetire != nil { |
| 289 | return a.RemoteRetire(ctx, target) |
| 290 | } |
| 291 | if !component(target.FenceID) { |
| 292 | return ErrTypedJournal |
| 293 | } |
| 294 | // Root's own stop inventory is the only authority for what this |
| 295 | // retirement may park. A saved journal is root's resumable progress in its |
| 296 | // own namespace; it can never add an artifact the inventory does not name. |
| 297 | in, err := a.Inventory.ReadExactStopInventory(ctx, target) |
| 298 | if err != nil { |
| 299 | return err |
| 300 | } |
| 301 | if !in.Complete || in.Box != target.Box || in.Enrollment != target.Enrollment || in.DirectoryRevision != target.DirectoryRevision { |
| 302 | return ErrTypedInventory |
| 303 | } |
| 304 | trustedRow, err := a.retirementRoute(row, target) |
| 305 | if err != nil { |
| 306 | return err |
| 307 | } |
| 308 | if err = ValidateAgentInventory(in, trustedRow, a.inventoryDataRoot()); err != nil { |
| 309 | return err |
| 310 | } |
| 311 | inventory := RetirementInventory{Host: row.Host, Captured: hostguard.Captured{Incarnation: row.ClaimIncarnation, Stamp: row.ClaimStamp}, Ownership: row.ClaimOwner} |
| 312 | for _, e := range in.Entries { |
| 313 | if e.Ownership != row.ClaimOwner || (e.Materialization != "materialized" && !(a.RequireTimerInventory && e.Kind == "timer" && e.Materialization == "intended")) { |
| 314 | return ErrTypedInventory |
| 315 | } |
| 316 | inventory.Artifacts = append(inventory.Artifacts, Artifact{AllowAbsent: e.Kind == "timer" && (e.TimerRetired || e.Materialization == "intended"), Kind: ArtifactKind(e.Kind), Root: e.ManagedRoot, Name: e.RelativePath, Owner: e.Ownership, RowID: e.RowIdentity, Identity: e.Identity, Dev: e.Device, Ino: e.Inode, Type: e.Type}) |
| 317 | } |
| 318 | if err = checkTrustedJournalDir(a.Retirement.JournalDir, true); err != nil { |
| 319 | return err |
| 320 | } |
| 321 | saved, loadErr := readTrustedJournal(filepath.Join(a.Retirement.JournalDir, target.FenceID+".json")) |
| 322 | if loadErr == nil { |
| 323 | var journal RetirementJournal |
| 324 | if decodeStrict(saved, &journal) != nil || journal.ID != target.FenceID || journal.Inventory.Host != row.Host || journal.Inventory.Captured != inventory.Captured || journal.Inventory.Ownership != row.ClaimOwner { |
| 325 | return ErrTypedJournal |
| 326 | } |
| 327 | // Timer rows retire in the journalled step, so the registry no longer |
| 328 | // lists them on resume. They are only ever this route's own registry |
| 329 | // rows; every other artifact must be exactly root's inventory. |
| 330 | var timers []Artifact |
| 331 | for _, x := range journal.Inventory.Artifacts { |
| 332 | if x.Kind == ArtifactTimer { |
| 333 | if x.Root != "registry" || x.Owner != row.ClaimOwner || !strings.HasPrefix(x.Name, row.Host+"/") || !managedName(strings.TrimPrefix(x.Name, row.Host+"/")) { |
| 334 | return ErrTypedJournal |
| 335 | } |
| 336 | timers = append(timers, x) |
| 337 | } |
| 338 | } |
| 339 | combined, err := retirementWithTimers(inventory, timers, a.RequireTimerInventory) |
| 340 | if err != nil { |
| 341 | return err |
| 342 | } |
| 343 | want, err := NewRetirementJournal(target.FenceID, combined) |
| 344 | if err != nil { |
| 345 | return ErrTypedJournal |
| 346 | } |
| 347 | got, err := NewRetirementJournal(target.FenceID, journal.Inventory) |
| 348 | if err != nil || !reflect.DeepEqual(want.Inventory, got.Inventory) { |
| 349 | return ErrTypedJournal |
| 350 | } |
| 351 | inventory = journal.Inventory |
| 352 | } else if !errors.Is(loadErr, os.ErrNotExist) { |
| 353 | return loadErr |
| 354 | } else { |
| 355 | timers, err := a.Registry.TimersFor(row.Host) |
| 356 | if err != nil { |
| 357 | return err |
| 358 | } |
| 359 | for _, t := range timers { |
| 360 | if t.ClaimIncarnation != row.ClaimIncarnation || t.ClaimStamp != row.ClaimStamp || t.ClaimOwner != row.ClaimOwner || t.ClaimRowID == "" || t.Registration == "" { |
| 361 | return ErrTypedInventory |
| 362 | } |
| 363 | inventory, err = retirementWithTimers(inventory, []Artifact{{Kind: ArtifactTimer, Root: "registry", Name: row.Host + "/" + t.Name, Owner: row.ClaimOwner, RowID: t.ClaimRowID + "~" + t.Registration}}, a.RequireTimerInventory) |
| 364 | if err != nil { |
| 365 | return err |
| 366 | } |
| 367 | } |
| 368 | } |
| 369 | |
| 370 | // FenceID is durable and unique for this exact stop and retirement scope. |
| 371 | intent, err := NewRetirementJournal(target.FenceID, inventory) |
| 372 | if err != nil { |
| 373 | return err |
| 374 | } |
| 375 | |
| 376 | receipt, err := a.Reconciler.Guard.FenceReceiptFor(ctx, target) |
| 377 | if err != nil { |
| 378 | return err |
| 379 | } |
| 380 | intent.Target = &target |
| 381 | intent.StopReceipt = append([]byte(nil), receipt.Bytes...) |
| 382 | retirement := a.Retirement |
| 383 | retirement.Rows = RegistryRows{Registry: a.Registry, Inventory: inventory} |
| 384 | return retirement.Run(ctx, target, intent) |
| 385 | } |
| 386 | |
| 387 | func (a ReconcileAdapter) inventoryDataRoot() string { |
| 388 | if a.DataRoot != "" { |
| 389 | return a.DataRoot |
| 390 | } |
| 391 | return a.Reconciler.Guard.DataDir |
| 392 | } |
| 393 | |
amadan