@volter/supercode-health 0.1.8 → 0.1.10

package/README.md

@@ -51,8 +51,9 @@ with the same outcome, the same expiry reason, ids and numbers taken out). A kin
 `windowMin` (60) raises `supercode.failing:<kind>`; it clears when the window holds none of it.
 
 `failures.json` in the health directory is the machine's ledger: the releases it ran, each from when it was first
-seen running, and when each kind was first seen and under which release. The ledger's first run records what is
-already failing as seeded, naming no release for it. A kind first seen under a release the ledger saw the machine
+seen running, and when each kind was first seen and under which release. The ledger's first run records every
+failure the machine's records hold (all of the slow log it keeps, every expiry), not only the window's, as seeded,
+naming no release for it; a run that could not read them seeds at the next. A kind first seen under a release the ledger saw the machine
 change to, after that change, is new with that release, and its alarm names that release's pull requests (the
 release writes them into `@volter/supercode`'s `lib/release-changes.json`). With `"failures": { "sayOnChanges":
 true }` in the config, the pack also comments such a kind once on each of those pull requests through `gh`, as the

package/lib/sample.mjs

@@ -3,6 +3,7 @@
 import { cpus, homedir, loadavg } from 'node:os';
 import { existsSync, readFileSync, statSync } from 'node:fs';
 import { run } from './run.mjs';
+import { daemonCall } from './mail.mjs';
 import { rendererOrigins, originOf, localPort } from './chrome.mjs';
 import { expiredFailures, failureSummary, keyFor, Ledger, runningRelease, sayOnChanges, slowFailures } from './failures.mjs';
 import { healthDir } from './paths.mjs';
@@ -423,7 +424,9 @@ export class Sampler {
         run(bin, ['teams', 'describe'], { timeoutMs: 15_000, env: this.env }).then((r) => ({ ...r, ms: Date.now() - t0 })),
         run(bin, ['teams', 'launches', 'list'], { timeoutMs: 45_000, env: this.env }),
         run(bin, ['teams', 'machines', 'list', '--json'], { timeoutMs: 20_000, env: this.env }),
-        run(bin, ['message', 'expired', '--json'], { timeoutMs: 20_000, env: this.env }),
+        // the mailbox's expiries are the daemon's own records (mail step 1): asked of it, never of the native binary
+        daemonCall('harness.v1.mail.expired', {}, { env: this.env, timeoutMs: 20_000 })
+          .then((answer) => ({ stdout: JSON.stringify(answer?.rows ?? []), error: null }), (error) => ({ stdout: '', error: error.message })),
       ]);
       let described = null; try { described = JSON.parse(describe.stdout); } catch { /* the daemon did not answer */ }
       let data = null; try { data = JSON.parse(status.stdout).data; } catch { /* unreadable */ }
@@ -474,6 +477,12 @@ export class Sampler {
       const release = runningRelease();
       this.ledger ??= new Ledger(join(healthDir(this.env), 'failures.json'));
       this.ledger.running(release, now);
+      // the ledger's first run seeds from all the records the machine keeps, not the window alone: a failure that comes
+      // and goes (once a few hours) was here before anything watched, and is never named new with a later release
+      if (this.ledger.fresh && !unread.length) {
+        for (const [kind, seen] of slowFailures(this.env, 0)) this.ledger.seen(kind, seen.last, release);
+        for (const [kind, seen] of expiredFailures(expiries, 0)) this.ledger.seen(kind, seen.last, release);
+      }
       const failing = [];
       for (const [kind, seen] of kinds) {
         const first = this.ledger.seen(kind, seen.last, release);
@@ -482,7 +491,8 @@ export class Sampler {
         slow.observations.push({ key: keyFor(kind), family: 'supercode', def: fdef, value: seen.count, unit: `in ${fdef.windowMin} min`, summary, detail: { kind, firstRelease: first.firstRelease, seeded: first.seeded, newWithRelease: isNew, running: release?.version ?? null, prs: isNew ? first.changes?.prs ?? [] : [] } });
         if (seen.count >= fdef.warn) failing.push({ kind, count: seen.count, first, summary });
       }
-      this.ledger.settled();
+      // settled only by a cycle that read both records: one that could not read them has not seeded the ledger
+      if (!unread.length) this.ledger.settled();
       slow.readings.supercode.failing = failing.map((f) => ({ kind: f.kind, count: f.count, firstRelease: f.first.firstRelease }));
       await sayOnChanges({ failing, machine: this.machine, release, ledger: this.ledger, env: this.env, config: cfg });
     });

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@volter/supercode-health",
-  "version": "0.1.8",
+  "version": "0.1.10",
   "type": "module",
   "description": "Optional machine pack: a cheap, deterministic probe of a machine's vitals that mails the machine's maintainer when a threshold is crossed. Each feature sits behind its own permission barrier.",
   "license": "MIT",