carlos / platform Public

Clone
git clone https://amadan.net/carlos/platform

Plain git — no account needed to clone.

Download

Download this file

1package host
2
3import (
4 "context"
5 "crypto/sha256"
6 "encoding/hex"
7 "encoding/json"
8 "errors"
9 "fmt"
10 "log"
11 "os"
12 "path/filepath"
13 "reflect"
14 "time"
15
16 "github.com/carlosframework/platform/internal/hostbox"
17 "github.com/carlosframework/platform/internal/hostclaim"
18 "github.com/carlosframework/platform/internal/hostguard"
19 "github.com/carlosframework/platform/internal/registry"
20 "github.com/carlosframework/platform/internal/release"
21)
22
23func (h ClaimRootHTTP) commitRootInventory(ctx context.Context, q hostbox.JournalRequest) (hostbox.JournalResponse, error) {
24 return h.withRootInventoryHistory(ctx, q, false)
25}
26func (h ClaimRootHTTP) withRootInventoryHistory(ctx context.Context, q hostbox.JournalRequest, pending bool) (hostbox.JournalResponse, error) {
27 if h.history == nil {
28 return h.commitRootInventoryPending(ctx, q, pending, nil)
29 }
30 var out hostbox.JournalResponse
31 err := h.history(ctx, func(history *release.RootHistory) error {
32 var err error
33 out, err = h.commitRootInventoryPending(ctx, q, pending, history)
34 return err
35 })
36 return out, err
37}
38
39func (h ClaimRootHTTP) commitRootInventoryPending(ctx context.Context, q hostbox.JournalRequest, pending bool, history *release.RootHistory) (hostbox.JournalResponse, error) {
40 var out hostbox.JournalResponse
41 if !pending && (q.V == 2 || q.V == 3) {
42 now := time.Now()
43 if h.Journals != nil && h.Journals.Now != nil {
44 now = h.Journals.Now()
45 }
46 age := now.Sub(time.Unix(0, q.IssuedAt))
47 if age > hostbox.JournalRequestWindow || age < -hostbox.JournalRequestWindow {
48 return out, hostbox.ErrJournalUncertain
49 }
50 }
51 guard := h.Stop.Coordinator.Guard
52 if guard.Root == nil || h.Journals == nil || h.InventoryActor == nil || h.Reconcile.Registry == nil || h.Reconcile.Census == nil {
53 return out, hostbox.ErrUnsupported
54 }
55 if h.RequireNewWork != nil {
56 if err := h.RequireNewWork(ctx); err != nil {
57 return out, err
58 }
59 }
60 actor, err := h.InventoryActor(ctx)
61 if err != nil {
62 return out, err
63 }
64 in, t := q.Inventory, q.Target
65 in.Entries = append([]hostbox.StopInventoryEntry(nil), in.Entries...)
66 // The inventory's stamp is validated against the fresh binding; a target
67 // stamp, when the agent sends one, must name that same binding before
68 // any intent is persisted (fix round 1, finding 2).
69 if q.Op != "inventory-write" || !in.Complete || in.Box != actor.Box || in.Enrollment != actor.Enrollment || in.DirectoryRevision != actor.Binding || (t.DirectoryRevision != "" && t.DirectoryRevision != in.DirectoryRevision) || t.Host != in.Host || t.Incarnation != in.Incarnation || t.Stamp != in.InstallationStamp || t.Box != actor.Box || t.Enrollment != actor.Enrollment || t.ClaimETag == "" || q.RecordKey == "" || q.RecordETag == "" {
70 return out, hostbox.ErrChanged
71 }
72 observed, err := hostclaim.Read(ctx, h.Stop.Coordinator.Evidence, in.Host)
73 if err != nil {
74 return out, err
75 }
76 if observed.Absent || observed.Claim == nil || observed.Claim.Holder == nil {
77 return out, hostbox.ErrChanged
78 }
79 applied := false
80 if pending {
81 saved, e := h.Journals.Inventory.SnapshotRequest(ctx, t)
82 if e != nil {
83 return out, e
84 }
85 applied = saved.Revision == q.Revision+1 && saved.LastRequest == q.ID
86 }
87 holder := observed.Claim.Holder
88 if (!applied && observed.ETag != t.ClaimETag) || holder.Incarnation != t.Incarnation || holder.Stamp != t.Stamp || holder.Session != t.Session || holder.Key != q.RecordKey || observed.Claim.Op.ID != t.Op || observed.Claim.Op.Phase != t.Phase {
89 return out, hostbox.ErrChanged
90 }
91 firstLanding := q.Revision == 0 && observed.Claim.Op.Kind == hostclaim.KindLanding
92 if firstLanding {
93 if err := rootFirstInstallation(observed.Claim, actor); err != nil {
94 return out, err
95 }
96 entry := rootLandingEntry(q, observed.Claim, actor)
97 if len(in.Entries) != 0 {
98 return out, hostbox.ErrTypedInventory
99 }
100 in.Entries = []hostbox.StopInventoryEntry{entry}
101 }
102 var canonical registry.Route
103 var currentRow *registry.Route
104 installation := hostguard.RootInstallation{Empty: len(in.Entries) == 0, Host: in.Host, Incarnation: in.Incarnation, Stamp: in.InstallationStamp, Migration: 1, InventoryRevision: q.Revision + 1, Account: holder.Account, App: holder.App, Record: holder.Key, Box: actor.Box, Enrollment: actor.Enrollment}
105 if q.Revision == 0 {
106 installation.Registration = hostguard.InstallationRegistration{Op: observed.Claim.Op.ID, Kind: observed.Claim.Op.Kind, Phase: observed.Claim.Op.Phase, Session: holder.Session}
107 if firstLanding {
108 installation.Registration.RowID = in.Entries[0].RowIdentity
109 }
110 }
111 raw, err := json.Marshal(q)
112 if err != nil {
113 return out, err
114 }
115 replayed := false
116 err = guard.CommitRootInventoryArtifacts(ctx, installation, raw, func(c *hostclaim.Claim) error {
117 if !reflect.DeepEqual(c, observed.Claim) {
118 return hostbox.ErrChanged
119 }
120 fresh, err := hostclaim.Read(ctx, h.Stop.Coordinator.Evidence, in.Host)
121 if err != nil {
122 return err
123 }
124 if !applied && fresh.ETag != t.ClaimETag {
125 return hostbox.ErrChanged
126 }
127 var etag string
128 canonical, etag, err = rootClaimRecord(ctx, h.Stop.Coordinator.Evidence, c, h.Roots["data"])
129 if err != nil {
130 return err
131 }
132 if etag != q.RecordETag {
133 return hostbox.ErrChanged
134 }
135 if c.Op.Kind == hostclaim.KindRelocating && q.Revision == 0 {
136 rec, e := relocationRecord(ctx, h.Roots["data"], h.Stop.Coordinator.Evidence, observed)
137 if e != nil {
138 return e
139 }
140 canonical = rec.Route
141 }
142 if c.Op.Kind == hostclaim.KindLanding {
143 canonical, err = h.rootLandingRoute(ctx, c, actor)
144 if err != nil {
145 return err
146 }
147 }
148 owner := sha256.Sum256([]byte(fmt.Sprintf("%s\x00%s\x00%d\x00%s", in.Host, in.Incarnation, in.InstallationStamp, actor.Box)))
149 canonical.ClaimIncarnation, canonical.ClaimStamp, canonical.ClaimOwner = in.Incarnation, in.InstallationStamp, hex.EncodeToString(owner[:])
150 for _, entry := range in.Entries {
151 if entry.Kind == "route" {
152 if canonical.ClaimRowID != "" {
153 return hostbox.ErrTypedInventory
154 }
155 canonical.ClaimRowID = entry.RowIdentity
156 }
157 }
158 row, err := h.Reconcile.Registry.Get(in.Host)
159 if err != nil {
160 return err
161 }
162 currentRow = row
163 if row != nil && (row.Host != canonical.Host || row.Account != canonical.Account || row.App != canonical.App || row.DBPath != canonical.DBPath || row.ClaimOwnership != canonical.ClaimOwnership) {
164 return hostbox.ErrTypedInventory
165 }
166 if row == nil {
167 for _, e := range in.Entries {
168 if e.Kind != "route" || (e.Materialization != "intended" && !firstLanding) {
169 return hostbox.ErrTypedInventory
170 }
171 }
172 } else if err := hostbox.ValidateAgentInventory(in, &canonical, h.Roots["data"]); err != nil {
173 return err
174 }
175 if err := h.verifyInventoryTimers(in); err != nil {
176 return err
177 }
178 if err := h.bindInventoryPaths(ctx, &in); err != nil {
179 return err
180 }
181 for _, e := range in.Entries {
182 if e.Ownership != canonical.ClaimOwner {
183 return hostbox.ErrTypedInventory
184 }
185 if e.Kind == "route" {
186 if e.ManagedRoot != "registry" || e.RelativePath != in.Host || e.RowIdentity == "" {
187 return hostbox.ErrTypedInventory
188 }
189 continue
190 }
191 if e.Kind == "timer" {
192 if err := h.verifyTimerMutationRow(in.Host, canonical.ClaimOwnership, e, q.Timer); err != nil {
193 return err
194 }
195 continue
196 }
197 if err := h.verifyInventoryPath(e); err != nil {
198 return err
199 }
200 }
201 return nil
202 }, func(priorRoot hostguard.RootInstallation) error {
203 if firstLanding && currentRow != nil && priorRoot.Registration != installation.Registration {
204 return hostbox.ErrTypedInventory
205 }
206 if firstLanding && currentRow == nil && priorRoot.InventoryRevision != 0 {
207 return hostbox.ErrTypedInventory
208 }
209 if observed.Claim.Op.Kind == hostclaim.KindLanding && !firstLanding && !landingRegistrationMatches(priorRoot, observed.Claim, currentRow) {
210 return hostbox.ErrTypedInventory
211 }
212
213 snapshot, err := h.Journals.Inventory.SnapshotRequest(ctx, hostguard.FenceTarget{Host: in.Host, Incarnation: in.Incarnation, Stamp: in.InstallationStamp})
214 prior, revision := snapshot.Inventory, snapshot.Revision
215 if err != nil {
216 return err
217 }
218 replay, found, err := h.Journals.Replay(ctx, actor.Box+"/"+actor.Enrollment, q)
219 if err != nil {
220 return err
221 }
222 if found {
223 if revision != q.Revision+1 || !reflect.DeepEqual(prior, in) {
224 return hostbox.ErrChanged
225 }
226 out = replay
227 replayed = true
228 return nil
229 }
230 if revision == q.Revision+1 && reflect.DeepEqual(prior, in) && (priorRoot.InventoryRevision == revision || pending && snapshot.LastRequest == q.ID) {
231 out = hostbox.JournalResponse{V: 1, ID: q.ID, Revision: revision}
232 replayed = true
233 return nil
234 }
235 if revision != q.Revision {
236 return hostbox.ErrChanged
237 }
238 if err := h.verifyNewInventoryIdentities(prior, in); err != nil {
239 return err
240 }
241 if err := h.verifyTimerMutation(prior, in, q.Timer); err != nil {
242 return err
243 }
244 if revision == 0 {
245 if (!firstLanding && len(in.Entries) != 0) || rootFirstInstallation(observed.Claim, actor) != nil {
246 return hostbox.ErrChanged
247 }
248 census := h.rootEmptyCensus()
249 selected := hostbox.RecordSelection{Host: in.Host, Key: holder.Key, ETag: q.RecordETag, Incarnation: in.Incarnation, Family: holder.Family}
250 if firstLanding && currentRow != nil && priorRoot.Registration == installation.Registration {
251 if err := census.CheckRootRegisteredRow(ctx, observed, selected, currentRow.ClaimOwnership); err != nil {
252 return err
253 }
254 } else if err := census.CheckLocal(ctx, observed, selected); err != nil {
255 return err
256 }
257 } else {
258 if priorRoot.InventoryRevision != revision {
259 return hostbox.ErrChanged
260 }
261 for _, old := range prior.Entries {
262 found := false
263 for _, next := range in.Entries {
264 if old.ArtifactID == next.ArtifactID {
265 if q.Timer != nil && old.Kind == "timer" && old.RelativePath == q.Timer.Timer.Host+"/"+q.Timer.Timer.Name {
266 old.RowIdentity, old.Materialization, old.TimerRetired = next.RowIdentity, next.Materialization, next.TimerRetired
267 }
268 if old.Materialization == "intended" && next.Materialization == "materialized" {
269 old.Materialization = next.Materialization
270 old.Identity = next.Identity
271 old.Device = next.Device
272 old.Inode = next.Inode
273 old.Type = next.Type
274 }
275 if old != next {
276 return hostbox.ErrChanged
277 }
278 found = true
279 }
280 }
281 if !found {
282 return hostbox.ErrChanged
283 }
284 }
285 }
286 return nil
287 }, func(prior hostguard.RootInstallation) error {
288 if history != nil && q.Revision == 0 && prior.InventoryRevision == 0 && observed.Claim.Op.Kind == hostclaim.KindServing && holder.Session == 0 && holder.WriterBox == "" {
289 if err := history.RecordCreation(release.HistoryBinding{Host: in.Host, Account: holder.Account, App: holder.App}, rootHistoryInstallation(installation)); err != nil {
290 return err
291 }
292 }
293 if firstLanding && currentRow == nil {
294 if err := h.Reconcile.Registry.Upsert(canonical); err != nil {
295 return err
296 }
297 }
298 if replayed {
299 return nil
300 }
301 if q.Timer != nil {
302 if err := h.applyTimerMutation(*q.Timer); err != nil {
303 return err
304 }
305 }
306 if firstLanding || h.inventoryIdentity != nil {
307 out, err = h.Journals.ExecuteRootInventory(ctx, actor.Box+"/"+actor.Enrollment, q, in, pending)
308 } else if pending {
309 out, err = h.Journals.CompletePendingInventory(ctx, actor.Box+"/"+actor.Enrollment, q)
310 } else {
311 out, err = h.Journals.Execute(ctx, actor.Box+"/"+actor.Enrollment, q)
312 }
313 return err
314 }, func(prior hostguard.RootInstallation) (hostguard.InstallationArtifacts, error) {
315 return rootInventoryArtifacts(prior, canonical, in, q.Revision)
316 })
317 return out, err
318}
319
320func (h ClaimRootHTTP) verifyInventoryPath(e hostbox.StopInventoryEntry) error {
321 if e.Kind != "path" || e.ManagedRoot != "data" || !filepath.IsLocal(e.RelativePath) {
322 return hostbox.ErrTypedInventory
323 }
324 path := filepath.Join(h.Roots["data"], e.RelativePath)
325 if e.Materialization == "intended" {
326 if _, err := os.Lstat(path); !errors.Is(err, os.ErrNotExist) {
327 return hostbox.ErrChanged
328 }
329 return nil
330 }
331 if e.Materialization != "materialized" {
332 return hostbox.ErrTypedInventory
333 }
334 return hostbox.VerifyInventoryIdentity(h.Roots["data"], e)
335}
336
337func (h ClaimRootHTTP) verifyNewInventoryIdentities(prior, next hostbox.StopInventory) error {
338 for _, e := range next.Entries {
339 if e.Kind != "path" || e.Materialization != "materialized" {
340 continue
341 }
342 retained := false
343 for _, old := range prior.Entries {
344 if old.ArtifactID == e.ArtifactID && old.Materialization == "materialized" {
345 retained = true
346 }
347 }
348 if retained {
349 continue
350 }
351 current, err := hostbox.CaptureInventoryIdentity(h.Roots[e.ManagedRoot], e.RelativePath)
352 if err != nil {
353 return err
354 }
355 if e.Identity.Boot != current.Boot {
356 return hostbox.ErrChanged
357 }
358 if err := h.verifyInventoryPath(e); err != nil {
359 return err
360 }
361 }
362 return nil
363}
364
365// obsoleteIntent reports whether a pending intent was stamped under a binding
366// other than the startup binding. Its stamp is validated binding evidence
367// only: an inventory's own stamp (checked against the fresh binding before
368// the intent was persisted) or a rebind's successor stamp (likewise). An
369// agent-sent target stamp is never consulted (fix round 1, finding 2).
370// Replaying an obsolete intent would compare the old binding with today's
371// authority; it is settled instead (N3).
372func (h ClaimRootHTTP) obsoleteIntent(stamp string) bool {
373 return h.startupBinding != "" && stamp != h.startupBinding
374}
375
376// settleObsoleteIntent settles an obsolete pending intent through the
377// guard's SettleObsoleteRootInventory (same host lock, exact intent bytes),
378// whatever the claim says now: the binding difference alone supersedes it.
379// The installation is quarantined and marked settled: it stays refused for
380// execution and new work, because every fresh comparison needs the new
381// binding, and the migration census at root's next start still inspects it,
382// so cleanup and migration startup are not blocked (fix round 1, finding 1).
383// applied says whether root's interrupted commit had already landed the
384// intent in the inventory; the guard then records the installation that
385// commit would have recorded, so guard and inventory agree. A settlement
386// failure is logged and leaves the intent pending, as any recovery failure
387// does. It never consults InventoryActor, so a quarantined box settles its
388// own leftovers.
389//
390// Cold-handover barriers (roster binding B3) are not pending installations
391// and are not settled here: RecoverPending's cold handover recovery settles
392// them (hostbox ColdHandover.Obsolete, hostguard
393// SettleObsoleteColdHandover) with the same startup binding.
394func (h ClaimRootHTTP) settleObsoleteIntent(ctx context.Context, host string, raw []byte, kind, stamped string, applied func(context.Context, hostguard.RootInstallation) (bool, error)) {
395 err := h.Stop.Coordinator.Guard.SettleObsoleteRootInventory(ctx, host, raw, applied)
396 log.Printf("instance-control: %s: obsolete pending %s intent: stamped %s current %s", host, kind, stamped, h.startupBinding)
397 if err != nil {
398 log.Printf("instance-control: %s: obsolete pending %s intent not settled: %v", host, kind, err)
399 }
400}
401
402// inventoryIntentApplied reports whether q, an interrupted inventory write,
403// already landed in the inventory: the pending installation's inventory is
404// at q's revision and names q as its last request, the test the pending
405// replay makes (commitRootInventoryPending's applied).
406func (h ClaimRootHTTP) inventoryIntentApplied(q hostbox.JournalRequest) func(context.Context, hostguard.RootInstallation) (bool, error) {
407 return func(ctx context.Context, pending hostguard.RootInstallation) (bool, error) {
408 if h.Journals == nil || h.Journals.Inventory == nil {
409 return false, hostbox.ErrUnsupported
410 }
411 saved, err := h.Journals.Inventory.SnapshotRequest(ctx, hostguard.FenceTarget{Host: pending.Host, Incarnation: pending.Incarnation, Stamp: pending.Stamp})
412 if err != nil {
413 return false, err
414 }
415 return saved.Revision == q.Revision+1 && saved.LastRequest == q.ID && saved.Revision == pending.InventoryRevision, nil
416 }
417}
418
419// rebindIntentUnapplied: a recovery rebind's commit writes only the
420// successor's inventory, so the predecessor the record keeps still agrees
421// with its own inventory; the record keeps it.
422func rebindIntentUnapplied(context.Context, hostguard.RootInstallation) (bool, error) {
423 return false, nil
424}
425
426func (h ClaimRootHTTP) recoverRootInventory(ctx context.Context) error {
427 authority := h.Stop.Coordinator.Guard.Root
428 if authority == nil {
429 return nil
430 }
431 pending, err := authority.PendingInstallations()
432 if err != nil {
433 return err
434 }
435 for _, raw := range pending {
436 var recovery recoveryRebindIntent
437 if err = json.Unmarshal(raw, &recovery); err != nil {
438 return err
439 }
440 if recovery.Recovery != nil {
441 if next := recovery.Recovery.Next; h.obsoleteIntent(next.DirectoryRevision) {
442 h.settleObsoleteIntent(ctx, next.Host, raw, "rebind", next.DirectoryRevision, rebindIntentUnapplied)
443 continue
444 }
445 if err = h.rebindRecoveryInventory(ctx, *recovery.Recovery); err != nil {
446 n := recovery.Recovery.Next
447 if errors.Is(err, hostbox.ErrChanged) {
448 err = errors.Join(err, h.Stop.Coordinator.Guard.AbandonRootInventory(ctx, n.Host, raw, func(c *hostclaim.Claim) bool {
449 fresh, e := hostclaim.Read(ctx, h.Stop.Coordinator.Evidence, n.Host)
450 return e == nil && fresh.Claim != nil && reflect.DeepEqual(c, fresh.Claim) && fresh.ETag != n.ETag
451 }, rebindIntentUnapplied))
452 }
453 log.Printf("instance-control: %s: %v", n.Host, err)
454 }
455 continue
456 }
457 var q hostbox.JournalRequest
458 if err = json.Unmarshal(raw, &q); err != nil {
459 return err
460 }
461 if settled, settleErr := h.settleTimerInventory(ctx, q, raw); settled {
462 continue
463 } else if settleErr != nil {
464 return settleErr
465 }
466 if h.obsoleteIntent(q.Inventory.DirectoryRevision) {
467 h.settleObsoleteIntent(ctx, q.Target.Host, raw, "inventory", q.Inventory.DirectoryRevision, h.inventoryIntentApplied(q))
468 continue
469 }
470 if _, err = h.withRootInventoryHistory(ctx, q, true); err != nil {
471 if errors.Is(err, hostbox.ErrChanged) || errors.Is(err, hostbox.ErrJournalUncertain) {
472 err = errors.Join(err, h.Stop.Coordinator.Guard.AbandonRootInventory(ctx, q.Target.Host, raw, func(c *hostclaim.Claim) bool {
473 fresh, readErr := hostclaim.Read(ctx, h.Stop.Coordinator.Evidence, q.Target.Host)
474 if readErr != nil || fresh.Claim == nil || !reflect.DeepEqual(c, fresh.Claim) {
475 return false
476 }
477 _, recordETag, recordErr := h.Stop.Coordinator.Evidence.Get(ctx, q.RecordKey)
478 return recordErr == nil && recordETag != q.RecordETag || fresh.ETag != q.Target.ClaimETag || c.Holder == nil || c.Holder.Incarnation != q.Target.Incarnation || c.Holder.Stamp != q.Target.Stamp || c.Holder.Session != q.Target.Session || c.Op.ID != q.Target.Op || c.Op.Phase != q.Target.Phase || errors.Is(err, hostbox.ErrJournalUncertain)
479 }, h.inventoryIntentApplied(q)))
480 }
481 log.Printf("instance-control: %s: %v", q.Target.Host, err)
482 }
483 }
484 return nil
485}
486