| 1 | package main |
| 2 | |
| 3 | // `carlos status --instances` — who is using what right now (discussion #11). |
| 4 | // |
| 5 | // The question "which account is using what on this box?" had no terminal |
| 6 | // answer. Every door holding it gated on a browser session, and the last |
| 7 | // person who needed it read four boxes' metrics objects and nine accounts' |
| 8 | // usage objects out of the bucket with `aws s3 cp` — which AGENTS.md's |
| 9 | // "operate via API and CLI" rule names as a product gap rather than a path. |
| 10 | // |
| 11 | // Console-backed, unlike `--live`. `--live` reads the status bucket with the |
| 12 | // operator's own bucket credentials, which is exactly the move this verb |
| 13 | // exists to make unnecessary: the console already sweeps the per-unit |
| 14 | // readings for its own browser page, and the account each unit belongs to is |
| 15 | // a registry join only the console can do. |
| 16 | |
| 17 | import ( |
| 18 | "context" |
| 19 | "encoding/json" |
| 20 | "errors" |
| 21 | "fmt" |
| 22 | "io" |
| 23 | "net/http" |
| 24 | "net/url" |
| 25 | "sort" |
| 26 | "strings" |
| 27 | ) |
| 28 | |
| 29 | // statusInstanceRow is one awake instance, as the console answers it |
| 30 | // (internal/console/apistatusinstances.go). Declared here rather than |
| 31 | // imported, the convention every other CLI-side wire struct follows. |
| 32 | // |
| 33 | // CPU and Mem are pointers because a missing reading is a real state — a unit |
| 34 | // seen for the first time has no CPU delta yet — and printing it as 0 would |
| 35 | // be printing a number an operator might act on. |
| 36 | type statusInstanceRow struct { |
| 37 | Box string `json:"box"` |
| 38 | Unit string `json:"unit"` |
| 39 | Host string `json:"host"` |
| 40 | Account string `json:"account"` |
| 41 | App string `json:"app"` |
| 42 | CPU *float64 `json:"cpu"` |
| 43 | Mem *int64 `json:"mem_bytes"` |
| 44 | Platform bool `json:"platform"` |
| 45 | } |
| 46 | |
| 47 | func (s *apiSession) statusInstanceList(ctx context.Context, account string) ([]statusInstanceRow, error) { |
| 48 | path := "/api/cli/status/instances" |
| 49 | if account != "" { |
| 50 | path += "?account=" + url.QueryEscape(account) |
| 51 | } |
| 52 | status, body, err := s.getJSON(ctx, path) |
| 53 | if err != nil { |
| 54 | return nil, err |
| 55 | } |
| 56 | if status != http.StatusOK { |
| 57 | if status == http.StatusNotFound { |
| 58 | // The same bare 404 an unknown route gives, and deliberately so: |
| 59 | // the answer names other people's accounts, so the refusal says |
| 60 | // nothing about them and nothing about who holds the operator |
| 61 | // bit. Which leaves this message covering both readings, because |
| 62 | // the caller genuinely cannot tell them apart from here. |
| 63 | return nil, errors.New("not found — this needs the fleet-operator bit on your account, " + |
| 64 | "or this console predates `carlos status --instances` (deploy a console built from current main)") |
| 65 | } |
| 66 | return nil, apiErrorFromBody(status, body) |
| 67 | } |
| 68 | var out struct { |
| 69 | Instances []statusInstanceRow `json:"instances"` |
| 70 | } |
| 71 | if err := json.Unmarshal(body, &out); err != nil { |
| 72 | return nil, fmt.Errorf("status instances: unreadable response: %w", err) |
| 73 | } |
| 74 | return out.Instances, nil |
| 75 | } |
| 76 | |
| 77 | func statusInstances(w io.Writer, consoleFlag, account string) error { |
| 78 | sess, err := resolveMode(consoleFlag) |
| 79 | if err != nil { |
| 80 | return err |
| 81 | } |
| 82 | if sess == nil { |
| 83 | return errors.New("carlos status --instances works through a console, and this shell is in direct bucket mode\n" + |
| 84 | "(CARLOS_DEPLOYMENT_BUCKET/CARLOS_DEPLOYMENT_DIR is set). Unset the store env and run:\n" + |
| 85 | "carlos auth login --console <url>") |
| 86 | } |
| 87 | rows, err := sess.statusInstanceList(context.Background(), account) |
| 88 | if err != nil { |
| 89 | return fmt.Errorf("status --instances: %w", err) |
| 90 | } |
| 91 | printStatusInstances(w, rows, account) |
| 92 | return nil |
| 93 | } |
| 94 | |
| 95 | func printStatusInstances(w io.Writer, rows []statusInstanceRow, account string) { |
| 96 | if len(rows) == 0 { |
| 97 | if account != "" { |
| 98 | mutedf(w, "no awake instances for %s right now", account) |
| 99 | } else { |
| 100 | mutedf(w, "no awake instances right now") |
| 101 | } |
| 102 | return |
| 103 | } |
| 104 | t := newTable(w) |
| 105 | fmt.Fprintln(t, "BOX\tHOST\tACCOUNT\tAPP\tCPU\tMEM") |
| 106 | for _, r := range rows { |
| 107 | host, app := r.Host, r.App |
| 108 | if r.Platform { |
| 109 | // A unit that matched no route: the console's own unit, the |
| 110 | // agent, system units. Named as what it is rather than left as |
| 111 | // two blank cells a reader would take for missing data — an |
| 112 | // operator asking what is on the box wants the box's own |
| 113 | // overhead in the picture. |
| 114 | host, app = r.Unit, "(platform)" |
| 115 | } |
| 116 | fmt.Fprintf(t, "%s\t%s\t%s\t%s\t%s\t%s\n", |
| 117 | r.Box, dashIfEmpty(host), dashIfEmpty(r.Account), dashIfEmpty(app), |
| 118 | cpuCell(r.CPU), memCell(r.Mem)) |
| 119 | } |
| 120 | t.Flush() |
| 121 | // Both said every time, because both are ways a short list misleads. |
| 122 | // |
| 123 | // A hibernating instance has no process, so it has no cgroup and no |
| 124 | // reading — it is not in this list and its absence is not evidence that |
| 125 | // it does not exist. The question this verb answers is "right now", and |
| 126 | // that limit belongs next to the answer. |
| 127 | mutedf(w, "(awake instances only — a hibernating instance has no process to measure)") |
| 128 | // The readings are fleet-wide, but the console joins them against its own |
| 129 | // box's route registry, so a unit on any other box matches nothing and |
| 130 | // lands under (platform) rather than under its tenant. Inert on a |
| 131 | // single-box deployment and fixed structurally by PR #137; until then an |
| 132 | // operator reading this for "who is using what" has to be told, the same |
| 133 | // way `carlos routes` says whose routes it is showing. |
| 134 | mutedf(w, "(accounts are joined against the console's own box's route registry; "+ |
| 135 | "an instance on another box in a fleet can land under (platform))") |
| 136 | } |
| 137 | |
| 138 | // statusLayoutRow is one host's layout, as the console answers it |
| 139 | // (internal/console/apistatuslayout.go). Declared here rather than |
| 140 | // imported, the same wire-struct convention statusInstanceRow above |
| 141 | // follows. |
| 142 | type statusLayoutRow struct { |
| 143 | Host string `json:"host"` |
| 144 | Layout string `json:"layout"` |
| 145 | Reason string `json:"reason"` |
| 146 | Note string `json:"note"` |
| 147 | } |
| 148 | |
| 149 | // statusLayoutBox is one box's layout report. |
| 150 | type statusLayoutBox struct { |
| 151 | Box string `json:"box"` |
| 152 | Mode string `json:"mode"` |
| 153 | Stale bool `json:"stale"` |
| 154 | Rows []statusLayoutRow `json:"rows"` |
| 155 | Mixed bool `json:"mixed"` |
| 156 | // Default: nested by the deployment's default, not a recorded intent; |
| 157 | // DefaultBlocked: why the default is not in force on this box. |
| 158 | Default bool `json:"default"` |
| 159 | DefaultBlocked string `json:"default_blocked"` |
| 160 | } |
| 161 | |
| 162 | func (s *apiSession) statusLayoutList(ctx context.Context) ([]statusLayoutBox, error) { |
| 163 | status, body, err := s.getJSON(ctx, "/api/cli/status/layout") |
| 164 | if err != nil { |
| 165 | return nil, err |
| 166 | } |
| 167 | if status != http.StatusOK { |
| 168 | if status == http.StatusNotFound { |
| 169 | // The same bare 404 an unknown route gives, and for the same |
| 170 | // reason statusInstanceList's own comment gives: the answer |
| 171 | // names every box's own layout, so the refusal must say |
| 172 | // nothing about whether the caller holds the operator bit. |
| 173 | return nil, errors.New("not found. this needs the fleet-operator bit on your account, " + |
| 174 | "or this console predates `carlos status --layout` (deploy a console built from current main)") |
| 175 | } |
| 176 | return nil, apiErrorFromBody(status, body) |
| 177 | } |
| 178 | var out struct { |
| 179 | Boxes []statusLayoutBox `json:"boxes"` |
| 180 | } |
| 181 | if err := json.Unmarshal(body, &out); err != nil { |
| 182 | return nil, fmt.Errorf("status layout: unreadable response: %w", err) |
| 183 | } |
| 184 | return out.Boxes, nil |
| 185 | } |
| 186 | |
| 187 | func statusLayout(w io.Writer, consoleFlag string) error { |
| 188 | sess, err := resolveMode(consoleFlag) |
| 189 | if err != nil { |
| 190 | return err |
| 191 | } |
| 192 | if sess == nil { |
| 193 | return errors.New("carlos status --layout works through a console, and this shell is in direct bucket mode\n" + |
| 194 | "(CARLOS_DEPLOYMENT_BUCKET/CARLOS_DEPLOYMENT_DIR is set). Unset the store env and run:\n" + |
| 195 | "carlos auth login --console <url>") |
| 196 | } |
| 197 | boxes, err := sess.statusLayoutList(context.Background()) |
| 198 | if err != nil { |
| 199 | return fmt.Errorf("status --layout: %w", err) |
| 200 | } |
| 201 | printStatusLayout(w, boxes) |
| 202 | return nil |
| 203 | } |
| 204 | |
| 205 | // printStatusLayout renders the door's answer as one row per host, sorted |
| 206 | // by box then host so two runs diff on the fleet rather than on request |
| 207 | // order — the door itself sorts, but a caller's own render must not depend |
| 208 | // on that staying true. The heading is the approved copy verbatim; rows are |
| 209 | // laid out to match it rather than through a tabwriter, whose dynamic |
| 210 | // column widths would drift the heading away from what was reviewed the |
| 211 | // moment any cell is wider than its header word. |
| 212 | func printStatusLayout(w io.Writer, boxes []statusLayoutBox) { |
| 213 | type line struct{ box, host, layout, reason string } |
| 214 | var lines []line |
| 215 | for _, b := range boxes { |
| 216 | name := b.Box |
| 217 | if b.Stale { |
| 218 | name += " (stale)" |
| 219 | } |
| 220 | // Nested by default is said on the box, not implied by its rows: |
| 221 | // an operator reading nested rows must be able to tell a box that |
| 222 | // was migrated on purpose from one that moved on its own. |
| 223 | if b.Default { |
| 224 | name += " (nested by default)" |
| 225 | } |
| 226 | for _, r := range b.Rows { |
| 227 | // A note is an event, not a refusal; it shares the column so |
| 228 | // the heading stays the approved one. |
| 229 | reason := r.Reason |
| 230 | if r.Note != "" { |
| 231 | if reason != "" { |
| 232 | reason += "; " |
| 233 | } |
| 234 | reason += r.Note |
| 235 | } |
| 236 | lines = append(lines, line{name, r.Host, r.Layout, reason}) |
| 237 | } |
| 238 | } |
| 239 | if len(lines) == 0 { |
| 240 | mutedf(w, "no boxes have reported a layout yet") |
| 241 | return |
| 242 | } |
| 243 | sort.Slice(lines, func(i, j int) bool { |
| 244 | if lines[i].box != lines[j].box { |
| 245 | return lines[i].box < lines[j].box |
| 246 | } |
| 247 | return lines[i].host < lines[j].host |
| 248 | }) |
| 249 | fmt.Fprintln(w, "BOX HOST LAYOUT REASON") |
| 250 | for _, l := range lines { |
| 251 | fmt.Fprintf(w, "%s %s %s %s\n", l.box, l.host, l.layout, dashIfEmpty(l.reason)) |
| 252 | } |
| 253 | // A box with flat, foreign or external rows beside nested ones, or |
| 254 | // under a nested intent, is the loud case: each of those still shares |
| 255 | // the agent's uid and mount namespace, so the nested rows on that box |
| 256 | // are not isolated from it yet. |
| 257 | for _, b := range boxes { |
| 258 | if b.Mixed { |
| 259 | fmt.Fprintf(w, "\nWARNING: %s runs nested instances beside flat, foreign or external ones, which share the agent's user and its view of /data. Nested isolation on this box is not in force until none of those remains.\n", b.Box) |
| 260 | } |
| 261 | } |
| 262 | // A box the default has not reached says why, once per box, so an |
| 263 | // operator waiting for a fleet to move on its own can see what each |
| 264 | // box is waiting for. |
| 265 | for _, b := range boxes { |
| 266 | if b.DefaultBlocked != "" && !b.Stale { |
| 267 | fmt.Fprintf(w, "\n%s is not nested by default: %s\n", b.Box, b.DefaultBlocked) |
| 268 | } |
| 269 | } |
| 270 | // Boxes that disagree about their mode are the other loud case: a |
| 271 | // box-scoped intent covers the boxes it names, and an instance placed |
| 272 | // or relocated onto a box outside it is minted flat there (#126, the |
| 273 | // Region HA 1b note). Said once, naming both sides. |
| 274 | if nested, other := layoutModeSplit(boxes); len(nested) > 0 && len(other) > 0 { |
| 275 | fmt.Fprintf(w, "\nWARNING: %s nested while %s not: an instance placed or moved onto a box that is not nested starts flat there. Run carlos system layout migrate without --box once every box is ready.\n", |
| 276 | boxList(nested), boxList(other)) |
| 277 | } |
| 278 | } |
| 279 | |
| 280 | // layoutModeSplit sorts the boxes by mode: nested, and anything else |
| 281 | // (flat, none, unreadable). Stale boxes are left out: their mode is what |
| 282 | // they last said, not what they are. |
| 283 | func layoutModeSplit(boxes []statusLayoutBox) (nested, other []string) { |
| 284 | for _, b := range boxes { |
| 285 | if b.Stale { |
| 286 | continue |
| 287 | } |
| 288 | if b.Mode == "nested" { |
| 289 | nested = append(nested, b.Box) |
| 290 | } else { |
| 291 | other = append(other, b.Box) |
| 292 | } |
| 293 | } |
| 294 | sort.Strings(nested) |
| 295 | sort.Strings(other) |
| 296 | return nested, other |
| 297 | } |
| 298 | |
| 299 | // boxList is "kass-1 is" or "kass-1 and kass-2 are", for a sentence. |
| 300 | func boxList(labels []string) string { |
| 301 | if len(labels) == 1 { |
| 302 | return labels[0] + " is" |
| 303 | } |
| 304 | return strings.Join(labels[:len(labels)-1], ", ") + " and " + labels[len(labels)-1] + " are" |
| 305 | } |
| 306 | |
| 307 | // cpuCell renders a fraction of one core. A nil reading prints as a dash: the |
| 308 | // unit was sampled but has no CPU delta yet, which is not the same as idle, |
| 309 | // and 0.000 is a claim about the process rather than about the sample. |
| 310 | func cpuCell(v *float64) string { |
| 311 | if v == nil { |
| 312 | return "—" |
| 313 | } |
| 314 | return fmt.Sprintf("%.3f", *v) |
| 315 | } |
| 316 | |
| 317 | // memCell renders bytes the way an operator scans them. Same nil rule. |
| 318 | func memCell(v *int64) string { |
| 319 | if v == nil { |
| 320 | return "—" |
| 321 | } |
| 322 | b := float64(*v) |
| 323 | switch { |
| 324 | case b >= 1<<30: |
| 325 | return fmt.Sprintf("%.1fGB", b/(1<<30)) |
| 326 | case b >= 1<<20: |
| 327 | return fmt.Sprintf("%.0fMB", b/(1<<20)) |
| 328 | case b >= 1<<10: |
| 329 | return fmt.Sprintf("%.0fKB", b/(1<<10)) |
| 330 | } |
| 331 | return fmt.Sprintf("%.0fB", b) |
| 332 | } |
| 333 | |
amadan