| 1 | package claimactivation |
| 2 | |
| 3 | import ( |
| 4 | "context" |
| 5 | "errors" |
| 6 | "os" |
| 7 | "reflect" |
| 8 | "sort" |
| 9 | "time" |
| 10 | |
| 11 | "github.com/carlosframework/platform/internal/hostactors" |
| 12 | "github.com/carlosframework/platform/internal/hostfloor" |
| 13 | "github.com/carlosframework/platform/internal/hostops" |
| 14 | "github.com/carlosframework/platform/internal/release" |
| 15 | ) |
| 16 | |
| 17 | // BackfillInput is the backfill phases' plan: the |
| 18 | // reservation id the boxes' census reports are published under |
| 19 | // (control/host-census/<id>/<box>.json). Nothing else is typed: the hosts |
| 20 | // come from the bucket's own record listing, and each host's box and region |
| 21 | // from the boxes' signed census reports. |
| 22 | type BackfillInput struct { |
| 23 | ID string `json:"id"` |
| 24 | } |
| 25 | |
| 26 | // BackfillHost is one frozen host, as read back from the signed plan. |
| 27 | type BackfillHost struct { |
| 28 | Host string `json:"host"` |
| 29 | hostops.BackfillHost |
| 30 | } |
| 31 | |
| 32 | // BackfillSummary is what the backfill phases read back from the signed |
| 33 | // plan after writing it (or, on a dry run, what they would write). |
| 34 | type BackfillSummary struct { |
| 35 | ID string `json:"id"` |
| 36 | Phase string `json:"phase"` |
| 37 | DirectoryRevision uint64 `json:"directory_revision"` |
| 38 | Hosts []BackfillHost `json:"hosts"` |
| 39 | // Reports is each box's census report digest, adopting only. |
| 40 | Reports map[string]string `json:"reports,omitempty"` |
| 41 | // Unowned lists the hosts no box reported. They keep no box and are |
| 42 | // never adopted. Adopting only. |
| 43 | Unowned []string `json:"unowned,omitempty"` |
| 44 | // Unreported lists the roster boxes with no census key, which |
| 45 | // backfill-done did not ask to report. Done only. |
| 46 | Unreported []string `json:"unreported,omitempty"` |
| 47 | } |
| 48 | |
| 49 | var ( |
| 50 | // ErrBackfillState is the refusal when the roster or the floor does not |
| 51 | // allow a backfill yet. |
| 52 | ErrBackfillState = errors.New("backfill refused: the roster must be published with a home region and a verified floor") |
| 53 | // ErrBackfillExists is backfill-collect's refusal over an existing plan. |
| 54 | ErrBackfillExists = errors.New("backfill refused: a backfill plan already exists") |
| 55 | // ErrBackfillRecords is backfill-collect's refusal when a record cannot |
| 56 | // be frozen: one without an acquisition, or records that changed while |
| 57 | // the plan was written. |
| 58 | ErrBackfillRecords = errors.New("backfill refused: every instance record must carry an acquisition and stay unchanged while the plan is written") |
| 59 | // ErrBackfillPlan is backfill-adopt's refusal when the plan is not the |
| 60 | // named reservation, still collecting, at this roster revision. |
| 61 | ErrBackfillPlan = errors.New("backfill refused: the plan must be the named reservation, still collecting, at this roster revision") |
| 62 | // ErrBackfillInstallations is backfill-done's refusal when a box's |
| 63 | // census report lists an installation: done is for an empty cohort, |
| 64 | // and there is no override. |
| 65 | ErrBackfillInstallations = errors.New("backfill refused: a census report lists installations; this phase is for an empty cohort") |
| 66 | ) |
| 67 | |
| 68 | // backfillReady is both phases' refusal up front: a published roster with a |
| 69 | // home region, whose floor digest is the published floor's. |
| 70 | func backfillReady(ctx context.Context, d hostactors.Directory, o hostactors.Observed) error { |
| 71 | if o.Roster.Publication != "published" || o.Roster.HomeRegion == "" || o.Roster.FloorSHA256 == "" { |
| 72 | return ErrBackfillState |
| 73 | } |
| 74 | floor, err := hostfloor.Read(ctx, d.Store, d.PublicKey, d.Bucket) |
| 75 | if err != nil { |
| 76 | return errors.Join(ErrBackfillState, err) |
| 77 | } |
| 78 | if floor.Digest != o.Roster.FloorSHA256 { |
| 79 | return ErrBackfillState |
| 80 | } |
| 81 | return nil |
| 82 | } |
| 83 | |
| 84 | // listBackfillHosts enumerates every record in the deployment through the |
| 85 | // same listing the box census and the adopters read (hostops.Apps and |
| 86 | // hostops.Records). |
| 87 | func listBackfillHosts(ctx context.Context, s release.Store) (map[string]hostops.BackfillHost, error) { |
| 88 | apps, err := hostops.Apps(ctx, s) |
| 89 | if err != nil { |
| 90 | return nil, err |
| 91 | } |
| 92 | hosts := map[string]hostops.BackfillHost{} |
| 93 | for _, app := range apps { |
| 94 | records, err := hostops.Records(ctx, s, app.Account, app.Name) |
| 95 | if err != nil { |
| 96 | return nil, err |
| 97 | } |
| 98 | for _, r := range records { |
| 99 | if _, dup := hosts[r.Holder.Host]; dup || r.Holder.Host == "" || r.Holder.Key == "" || r.ETag == "" { |
| 100 | return nil, ErrBackfillRecords |
| 101 | } |
| 102 | if r.Holder.Acquisition == "" { |
| 103 | return nil, ErrBackfillRecords |
| 104 | } |
| 105 | hosts[r.Holder.Host] = hostops.BackfillHost{Key: r.Holder.Key, ETag: r.ETag, Acquisition: r.Holder.Acquisition} |
| 106 | } |
| 107 | } |
| 108 | return hosts, nil |
| 109 | } |
| 110 | |
| 111 | func backfillSummary(p hostops.BackfillPlan) BackfillSummary { |
| 112 | out := BackfillSummary{ID: p.ID, Phase: p.Phase, DirectoryRevision: p.DirectoryRevision, Hosts: []BackfillHost{}, Reports: p.Reports} |
| 113 | for host, h := range p.Hosts { |
| 114 | out.Hosts = append(out.Hosts, BackfillHost{Host: host, BackfillHost: h}) |
| 115 | if p.Phase == "adopting" && h.Box == "" { |
| 116 | out.Unowned = append(out.Unowned, host) |
| 117 | } |
| 118 | } |
| 119 | sort.Slice(out.Hosts, func(i, j int) bool { return out.Hosts[i].Host < out.Hosts[j].Host }) |
| 120 | sort.Strings(out.Unowned) |
| 121 | return out |
| 122 | } |
| 123 | |
| 124 | // readSignedPlan reads the backfill plan and verifies its root signature. |
| 125 | func readSignedPlan(ctx context.Context, d hostactors.Directory) (hostops.BackfillPlan, string, error) { |
| 126 | var p hostops.BackfillPlan |
| 127 | raw, tag, err := d.Store.Get(ctx, hostops.BackfillKey) |
| 128 | if err != nil { |
| 129 | return p, "", err |
| 130 | } |
| 131 | if tag == "" || release.VerifyRootEnvelope(d.PublicKey, raw, &p) != nil || p.V != 1 { |
| 132 | return p, "", ErrBackfillPlan |
| 133 | } |
| 134 | return p, tag, nil |
| 135 | } |
| 136 | |
| 137 | // backfillCollect writes the signed collecting plan, create-only. The plan |
| 138 | // freezes every instance record in the deployment by key and etag: the box |
| 139 | // census refuses any record it does not name, and claim-mode record writers |
| 140 | // refuse while it exists. It is read back, verified, and the records are |
| 141 | // listed again afterwards: a record that moved while the plan was written |
| 142 | // is a refusal, because the box census would refuse the plan anyway. |
| 143 | func (r Runner) backfillCollect(ctx context.Context, d hostactors.Directory, o hostactors.Observed, in BackfillInput, result *Result) error { |
| 144 | if !validBackfillID(in.ID) { |
| 145 | return ErrActivation |
| 146 | } |
| 147 | if err := backfillReady(ctx, d, o); err != nil { |
| 148 | return err |
| 149 | } |
| 150 | if _, _, err := d.Store.Get(ctx, hostops.BackfillKey); err == nil { |
| 151 | return ErrBackfillExists |
| 152 | } else if !errors.Is(err, os.ErrNotExist) { |
| 153 | return err |
| 154 | } |
| 155 | hosts, err := listBackfillHosts(ctx, d.Store) |
| 156 | if err != nil { |
| 157 | return err |
| 158 | } |
| 159 | p := hostops.BackfillPlan{V: 1, ID: in.ID, Phase: "collecting", DirectoryRevision: o.Roster.Revision, CreatedAt: time.Now().UTC(), Hosts: hosts} |
| 160 | body, err := release.SignRootEnvelope(r.Signer, p) |
| 161 | if err != nil { |
| 162 | return err |
| 163 | } |
| 164 | if _, err = d.Store.Put(ctx, hostops.BackfillKey, body, ""); err != nil { |
| 165 | return err |
| 166 | } |
| 167 | got, _, err := readSignedPlan(ctx, d) |
| 168 | if err != nil { |
| 169 | return err |
| 170 | } |
| 171 | if got.Phase != "collecting" || got.ID != in.ID || got.DirectoryRevision != o.Roster.Revision || !reflect.DeepEqual(got.Hosts, hosts) { |
| 172 | return ErrBackfillPlan |
| 173 | } |
| 174 | again, err := listBackfillHosts(ctx, d.Store) |
| 175 | if err != nil { |
| 176 | return err |
| 177 | } |
| 178 | if !reflect.DeepEqual(again, hosts) { |
| 179 | return ErrBackfillRecords |
| 180 | } |
| 181 | summary := backfillSummary(got) |
| 182 | result.Backfill = &summary |
| 183 | return nil |
| 184 | } |
| 185 | |
| 186 | // backfillAdopt advances the named collecting plan to adopting through |
| 187 | // hostops' Backfill.Publish, under the plan's own etag (If-Match). Publish |
| 188 | // requires every roster box's signed census report for the reservation, |
| 189 | // freezes their digests, takes each host's box and region from them, and |
| 190 | // refuses unfinished seed state. The adopters' own acceptance |
| 191 | // (VerifyAdopting) is then run on the result as readback. |
| 192 | func (r Runner) backfillAdopt(ctx context.Context, d hostactors.Directory, o hostactors.Observed, in BackfillInput, result *Result) error { |
| 193 | if !validBackfillID(in.ID) { |
| 194 | return ErrActivation |
| 195 | } |
| 196 | if err := backfillReady(ctx, d, o); err != nil { |
| 197 | return err |
| 198 | } |
| 199 | p, tag, err := readSignedPlan(ctx, d) |
| 200 | if err != nil { |
| 201 | return err |
| 202 | } |
| 203 | if p.ID != in.ID || p.Phase != "collecting" || p.DirectoryRevision != o.Roster.Revision { |
| 204 | return ErrBackfillPlan |
| 205 | } |
| 206 | b := &hostops.Backfill{Store: d.Store, Directory: d} |
| 207 | if err := b.Publish(ctx, r.Signer, tag); err != nil { |
| 208 | return err |
| 209 | } |
| 210 | got, err := b.VerifyAdopting(ctx) |
| 211 | if err != nil { |
| 212 | return err |
| 213 | } |
| 214 | if got.ID != in.ID || got.DirectoryRevision != o.Roster.Revision || len(got.Hosts) != len(p.Hosts) { |
| 215 | return ErrBackfillPlan |
| 216 | } |
| 217 | for host, h := range p.Hosts { |
| 218 | n, ok := got.Hosts[host] |
| 219 | if !ok || n.Key != h.Key || n.ETag != h.ETag || n.Acquisition != h.Acquisition { |
| 220 | return ErrBackfillPlan |
| 221 | } |
| 222 | } |
| 223 | summary := backfillSummary(got) |
| 224 | result.Backfill = &summary |
| 225 | return nil |
| 226 | } |
| 227 | |
| 228 | // backfillDone closes the named collecting plan straight to done, for an |
| 229 | // empty cohort (claim trial Option B, Paul's D1 of 2026-10-20), through |
| 230 | // hostops' Backfill.Done under the plan's own etag (If-Match). Done adopts |
| 231 | // nothing: it refuses unless the records are the plan's, no host claim or |
| 232 | // app admission exists, and every roster box with a census key reports no |
| 233 | // installation. Boxes with no key are listed as unreported. The plan is then |
| 234 | // read back as done. |
| 235 | func (r Runner) backfillDone(ctx context.Context, d hostactors.Directory, o hostactors.Observed, in BackfillInput, result *Result) error { |
| 236 | if !validBackfillID(in.ID) { |
| 237 | return ErrActivation |
| 238 | } |
| 239 | if err := backfillReady(ctx, d, o); err != nil { |
| 240 | return err |
| 241 | } |
| 242 | p, tag, err := readSignedPlan(ctx, d) |
| 243 | if err != nil { |
| 244 | return err |
| 245 | } |
| 246 | if p.ID != in.ID || p.Phase != "collecting" || p.DirectoryRevision != o.Roster.Revision { |
| 247 | return ErrBackfillPlan |
| 248 | } |
| 249 | b := &hostops.Backfill{Store: d.Store, Directory: d} |
| 250 | unreported, err := b.Done(ctx, r.Signer, tag) |
| 251 | switch { |
| 252 | case errors.Is(err, hostops.ErrDoneInstallations): |
| 253 | return errors.Join(ErrBackfillInstallations, err) |
| 254 | case errors.Is(err, hostops.ErrDoneRecords): |
| 255 | return errors.Join(ErrBackfillRecords, err) |
| 256 | case errors.Is(err, hostops.ErrDoneETag), errors.Is(err, hostops.ErrDoneEnvelope), errors.Is(err, hostops.ErrDonePhase): |
| 257 | return errors.Join(ErrBackfillPlan, err) |
| 258 | case err != nil: |
| 259 | return err |
| 260 | } |
| 261 | got, _, err := readSignedPlan(ctx, d) |
| 262 | if err != nil { |
| 263 | return err |
| 264 | } |
| 265 | if got.ID != in.ID || got.Phase != "done" || got.DirectoryRevision != o.Roster.Revision || len(got.Hosts) != len(p.Hosts) { |
| 266 | return ErrBackfillPlan |
| 267 | } |
| 268 | for host, h := range p.Hosts { |
| 269 | if got.Hosts[host] != h { |
| 270 | return ErrBackfillPlan |
| 271 | } |
| 272 | } |
| 273 | summary := backfillSummary(got) |
| 274 | summary.Unreported = unreported |
| 275 | result.Backfill = &summary |
| 276 | return nil |
| 277 | } |
| 278 | |
| 279 | // validBackfillID keeps the reservation id one safe path segment. |
| 280 | func validBackfillID(id string) bool { |
| 281 | if id == "" || len(id) > 64 || id == "." || id == ".." { |
| 282 | return false |
| 283 | } |
| 284 | for _, c := range id { |
| 285 | if !(c >= 'a' && c <= 'z' || c >= '0' && c <= '9' || c == '-' || c == '_' || c == '.') { |
| 286 | return false |
| 287 | } |
| 288 | } |
| 289 | return true |
| 290 | } |
| 291 | |
amadan