| 1 | package host |
| 2 | |
| 3 | import ( |
| 4 | "context" |
| 5 | "encoding/base64" |
| 6 | "encoding/json" |
| 7 | "errors" |
| 8 | "fmt" |
| 9 | "io" |
| 10 | "log" |
| 11 | "net/http" |
| 12 | "os" |
| 13 | "strconv" |
| 14 | "strings" |
| 15 | "time" |
| 16 | |
| 17 | "github.com/carlosframework/platform/internal/boxid" |
| 18 | "github.com/carlosframework/platform/internal/release" |
| 19 | ) |
| 20 | |
| 21 | // regionHAEnabled is CARLOS_REGION_HA: on only for a value |
| 22 | // strconv.ParseBool reads as true ("1", "t", "true" in any case); "", |
| 23 | // "0", "false", "off", "no" and anything unparseable are off. Every region |
| 24 | // HA behaviour on a box is behind it (spec 1 §8 step 2). |
| 25 | func regionHAEnabled(getenv func(string) string) bool { |
| 26 | on, err := strconv.ParseBool(strings.TrimSpace(getenv("CARLOS_REGION_HA"))) |
| 27 | return err == nil && on |
| 28 | } |
| 29 | |
| 30 | const ( |
| 31 | imdsBase = "http://169.254.169.254" |
| 32 | imdsIdentityDocument = "/latest/dynamic/instance-identity/document" |
| 33 | imdsIdentitySignature = "/latest/dynamic/instance-identity/signature" |
| 34 | boxRecordRetry = time.Minute |
| 35 | ) |
| 36 | |
| 37 | // imdsGetRaw reads one IMDSv2 path and returns its body untouched. The |
| 38 | // identity document is signed byte for byte, so unlike imdsLeaf nothing is |
| 39 | // trimmed. |
| 40 | func imdsGetRaw(client *http.Client, base, token, path string) ([]byte, error) { |
| 41 | req, err := http.NewRequest(http.MethodGet, base+path, nil) |
| 42 | if err != nil { |
| 43 | return nil, err |
| 44 | } |
| 45 | req.Header.Set("X-aws-ec2-metadata-token", token) |
| 46 | resp, err := client.Do(req) |
| 47 | if err != nil { |
| 48 | return nil, err |
| 49 | } |
| 50 | defer resp.Body.Close() |
| 51 | body, err := io.ReadAll(io.LimitReader(resp.Body, 64<<10)) |
| 52 | if err != nil { |
| 53 | return nil, err |
| 54 | } |
| 55 | if resp.StatusCode != http.StatusOK { |
| 56 | return nil, fmt.Errorf("IMDS %s answered %d", path, resp.StatusCode) |
| 57 | } |
| 58 | return body, nil |
| 59 | } |
| 60 | |
| 61 | // readIdentityRecord builds this box's record (spec §5.1) from IMDS: the |
| 62 | // identity document exactly as served, and its base64 RSA-SHA256 |
| 63 | // signature with any line breaks removed (base64 ignores them; Go's |
| 64 | // decoder does not). |
| 65 | func readIdentityRecord(client *http.Client, base, box string, now time.Time) (boxid.Record, error) { |
| 66 | req, err := http.NewRequest(http.MethodPut, base+"/latest/api/token", nil) |
| 67 | if err != nil { |
| 68 | return boxid.Record{}, err |
| 69 | } |
| 70 | req.Header.Set("X-aws-ec2-metadata-token-ttl-seconds", "60") |
| 71 | resp, err := client.Do(req) |
| 72 | if err != nil { |
| 73 | return boxid.Record{}, fmt.Errorf("IMDS token: %w", err) |
| 74 | } |
| 75 | tok, _ := io.ReadAll(io.LimitReader(resp.Body, 4<<10)) |
| 76 | resp.Body.Close() |
| 77 | if resp.StatusCode != http.StatusOK || len(strings.TrimSpace(string(tok))) == 0 { |
| 78 | return boxid.Record{}, fmt.Errorf("IMDS token answered %d", resp.StatusCode) |
| 79 | } |
| 80 | token := strings.TrimSpace(string(tok)) |
| 81 | doc, err := imdsGetRaw(client, base, token, imdsIdentityDocument) |
| 82 | if err != nil { |
| 83 | return boxid.Record{}, err |
| 84 | } |
| 85 | sig, err := imdsGetRaw(client, base, token, imdsIdentitySignature) |
| 86 | if err != nil { |
| 87 | return boxid.Record{}, err |
| 88 | } |
| 89 | cleaned := strings.Join(strings.Fields(string(sig)), "") |
| 90 | if _, err := base64.StdEncoding.DecodeString(cleaned); err != nil || cleaned == "" { |
| 91 | return boxid.Record{}, errors.New("IMDS signature is not base64") |
| 92 | } |
| 93 | return boxid.Record{V: 1, Box: box, IdentityDocument: base64.StdEncoding.EncodeToString(doc), |
| 94 | IdentitySignature: cleaned, WrittenAt: now.UTC().Format(time.RFC3339)}, nil |
| 95 | } |
| 96 | |
| 97 | // recordInstanceID is the instanceId in a record's identity document. It |
| 98 | // does not check the signature: it only tells whether a record names this |
| 99 | // box's instance. |
| 100 | func recordInstanceID(rec boxid.Record) (string, error) { |
| 101 | doc, err := base64.StdEncoding.DecodeString(rec.IdentityDocument) |
| 102 | if err != nil { |
| 103 | return "", errors.New("identity document is not base64") |
| 104 | } |
| 105 | var d struct { |
| 106 | InstanceID string `json:"instanceId"` |
| 107 | } |
| 108 | if err := json.Unmarshal(doc, &d); err != nil || d.InstanceID == "" { |
| 109 | return "", errors.New("identity document names no instance") |
| 110 | } |
| 111 | return d.InstanceID, nil |
| 112 | } |
| 113 | |
| 114 | // publishBoxRecord makes one attempt to publish this box's enrollment |
| 115 | // evidence. done is true once there is nothing left to retry: the record |
| 116 | // is in place, or a different one already holds this reference, which |
| 117 | // only an operator can sort out and which is logged loudly. A record |
| 118 | // already under this reference is never rewritten, but it is checked |
| 119 | // against this box's own identity: one signed for another instance, or |
| 120 | // one that cannot be read, is reported the same loud way. |
| 121 | func publishBoxRecord(ctx context.Context, s release.Store, box string, read func() (boxid.Record, error)) (done bool, err error) { |
| 122 | if !boxid.ValidBox(box) { |
| 123 | return true, fmt.Errorf("box reference %q is not a label; no enrollment evidence is published", box) |
| 124 | } |
| 125 | existing, err := boxid.Load(ctx, s, box) |
| 126 | if err != nil && !errors.Is(err, os.ErrNotExist) { |
| 127 | return false, err |
| 128 | } |
| 129 | rec, rerr := read() |
| 130 | if rerr != nil { |
| 131 | return false, rerr |
| 132 | } |
| 133 | if err == nil { |
| 134 | var held boxid.Record |
| 135 | if jerr := json.Unmarshal(existing, &held); jerr != nil { |
| 136 | return true, fmt.Errorf("the record already under %s cannot be read: %v", box, jerr) |
| 137 | } |
| 138 | theirs, ierr := recordInstanceID(held) |
| 139 | if ierr != nil { |
| 140 | return true, fmt.Errorf("the record already under %s: %v", box, ierr) |
| 141 | } |
| 142 | mine, ierr := recordInstanceID(rec) |
| 143 | if ierr != nil { |
| 144 | return false, fmt.Errorf("this box's own identity: %v", ierr) |
| 145 | } |
| 146 | if theirs != mine { |
| 147 | return true, fmt.Errorf("the record already under %s is signed for instance %s, this box is %s", box, theirs, mine) |
| 148 | } |
| 149 | return true, nil // create-only: an existing record is never rewritten |
| 150 | } |
| 151 | err = boxid.Publish(ctx, s, rec) |
| 152 | if errors.Is(err, release.ErrCASConflict) { |
| 153 | return true, fmt.Errorf("another record already holds %s: %w", box, err) |
| 154 | } |
| 155 | return err == nil, err |
| 156 | } |
| 157 | |
| 158 | // boxRecordLoop publishes this box's enrollment evidence once, at start, |
| 159 | // when CARLOS_REGION_HA is on, retrying each minute until it is in place. |
| 160 | // The record grants nothing: enrollment (an operator, or for Auto Scaling |
| 161 | // boxes the console with a membership confirmation) reads it later. |
| 162 | // Platform-box job: skipped on fleet boxes. |
| 163 | func boxRecordLoop() { |
| 164 | if !regionHAEnabled(os.Getenv) || shouldSkipPublisherElection(os.Getenv) { |
| 165 | return |
| 166 | } |
| 167 | ctx := context.Background() |
| 168 | store, err := openDeployStore(ctx) |
| 169 | if err != nil { |
| 170 | log.Printf("agent: box record not published: %v", err) |
| 171 | return |
| 172 | } |
| 173 | client := &http.Client{Timeout: 2 * time.Second} |
| 174 | for attempt := 1; ; attempt++ { |
| 175 | // A group box's reference comes from its identity document, which |
| 176 | // IMDS may not answer yet: that is a retry, never a refusal |
| 177 | // (boxreference.go). |
| 178 | box := metricsBoxLabel(os.Getenv, readMachineID()) |
| 179 | read := func() (boxid.Record, error) { return readIdentityRecord(client, imdsBase, box, time.Now()) } |
| 180 | done, err := false, errors.New("this box's group reference could not be read") |
| 181 | if box != "" { |
| 182 | done, err = publishBoxRecord(ctx, store, box, read) |
| 183 | } |
| 184 | name := box |
| 185 | if name == "" { |
| 186 | name = "this group box" |
| 187 | } |
| 188 | if done { |
| 189 | if err != nil { |
| 190 | log.Printf("agent: box record for %s: %v (region HA cannot enroll this box until an operator resolves it)", name, err) |
| 191 | } |
| 192 | return |
| 193 | } |
| 194 | if attempt == 1 || attempt%60 == 0 { |
| 195 | log.Printf("agent: box record for %s not published yet (#%d): %v", name, attempt, err) |
| 196 | } |
| 197 | time.Sleep(boxRecordRetry) |
| 198 | } |
| 199 | } |
| 200 | |
amadan