fix(relay): let the rehome evidence parser read a line the director grew (#21823)

The enable workflow reads the director's `[orca-relay] regional rehome inventory` line out of Cloud Logging and pins the whole line with one regex. Adding `hostNotArrivedLast24Hours` in #21813 made every healthy line stop matching, so "Read fresh aggregate completion and abort evidence" threw "no aggregate regional rehome inventory evidence" and the fail-closed step disabled the durable switch at control generation 26. The parser now requires the six original fields and tolerates further ones in any order. Extra fields stay fenced by value shape rather than by pinning the whole line: a field must be a bare name and a non-negative integer or `none`, so `hostId=someone` is still not a counter and cannot ride along. An absent count reads as null, not zero, because an older director not reporting leaks is not the same as reporting none. `hostNotArrivedLast24Hours` and `oldestActiveAgeMs` now reach the evidence JSON and the operator step summary. Two guards close the chain, each verified to fail on the regression it exists for: a census in the relay package feeds the real formatter's output to the real parser, and a script-side test pins the parser's output to the fields the workflow summary renders. Claude-Session: https://claude.ai/session/ced32ebb-7155-4413-adad-1eccd14c2010
2026-09-22 00:02:31 +00:00 · 2026-09-20 15:50:43 -04:00
parent fa0010e8d6
commit 68b11282a5
4 changed files with 311 additions and 10 deletions
@@ -1,6 +1,24 @@
 import { pathToFileURL } from 'node:url'

-const INVENTORY = /^\[orca-relay\] regional rehome inventory active=(\d+) awaitingReceipt=(\d+) targetRegistered=(\d+) completedLast24Hours=(\d+) abortedLast24Hours=(\d+) oldestActiveAgeMs=(none|\d+)$/
+const INVENTORY_PREFIX = '[orca-relay] regional rehome inventory '
+// A counters line, so every field is a bare name and a non-negative integer or
+// `none`. Pinning the whole line instead is what broke the enable workflow when
+// `hostNotArrivedLast24Hours` shipped: the director grew a field and the parser
+// read a healthy line as no evidence at all. Tolerating extra fields is safe
+// only because the value shape stays fenced — `hostId=someone` is still not a
+// counter, so an identity-bearing lookalike cannot slip through as an extra.
+const FIELD = /^([A-Za-z][A-Za-z0-9]*)=(none|\d{1,15})$/
+// `oldestActiveAgeMs` is the one required field the director can report as
+// `none`; a count that reads `none` is a line this parser does not recognise,
+// not evidence worth failing the run over.
+const REQUIRED_COUNTS = [
+  'active',
+  'awaitingReceipt',
+  'targetRegistered',
+  'completedLast24Hours',
+  'abortedLast24Hours'
+]
+const REQUIRED_FIELDS = [...REQUIRED_COUNTS, 'oldestActiveAgeMs']

 function count(value, name) {
  const parsed = Number(value)
@@ -8,20 +26,43 @@ function count(value, name) {
  return parsed
 }

+// Returns the field map, or null for anything that is not this line.
+export function readRegionalRehomeInventoryFields(textPayload) {
+  if (typeof textPayload !== 'string' || !textPayload.startsWith(INVENTORY_PREFIX)) return null
+  const fields = new Map()
+  for (const token of textPayload.slice(INVENTORY_PREFIX.length).split(' ')) {
+    const field = FIELD.exec(token)
+    if (!field || fields.has(field[1])) return null
+    fields.set(field[1], field[2])
+  }
+  if (!REQUIRED_FIELDS.every((name) => fields.has(name))) return null
+  if (REQUIRED_COUNTS.some((name) => fields.get(name) === 'none')) return null
+  return fields
+}
+
+// Absent is not zero: a director on an older image emits no such field, and
+// reporting 0 would read as "no leaks" rather than "not measured".
+function optionalCount(fields, name) {
+  const value = fields.get(name)
+  if (value === undefined || value === 'none') return null
+  return count(value, name)
+}
+
 export function parseRegionalRehomeInventory(entries, options = {}) {
  if (!Array.isArray(entries)) throw new Error('logging response must be an array')
  const parsed = entries.flatMap((entry) => {
-    const match = INVENTORY.exec(entry?.textPayload ?? '')
+    const fields = readRegionalRehomeInventoryFields(entry?.textPayload ?? '')
    const timestamp = Date.parse(entry?.timestamp ?? '')
-    if (!match || !Number.isFinite(timestamp)) return []
+    if (!fields || !Number.isFinite(timestamp)) return []
    return [{
      timestamp,
-      active: count(match[1], 'active'),
-      awaitingReceipt: count(match[2], 'awaiting receipt'),
-      targetRegistered: count(match[3], 'target registered'),
-      completedLast24Hours: count(match[4], 'completed'),
-      abortedLast24Hours: count(match[5], 'aborted'),
-      oldestActiveAgeMs: match[6] === 'none' ? null : count(match[6], 'oldest active age')
+      active: count(fields.get('active'), 'active'),
+      awaitingReceipt: count(fields.get('awaitingReceipt'), 'awaiting receipt'),
+      targetRegistered: count(fields.get('targetRegistered'), 'target registered'),
+      completedLast24Hours: count(fields.get('completedLast24Hours'), 'completed'),
+      abortedLast24Hours: count(fields.get('abortedLast24Hours'), 'aborted'),
+      hostNotArrivedLast24Hours: optionalCount(fields, 'hostNotArrivedLast24Hours'),
+      oldestActiveAgeMs: optionalCount(fields, 'oldestActiveAgeMs')
    }]
  }).sort((left, right) => right.timestamp - left.timestamp)
  if (parsed.length === 0) throw new Error('no aggregate regional rehome inventory evidence')
@@ -1,6 +1,7 @@
 import assert from 'node:assert/strict'
 import { test } from 'node:test'
 import { parseRegionalRehomeInventory } from './relay-rehome-aggregate-evidence.mjs'
+import { readRelayWorkflow } from './relay-repository.mjs'

 const now = Date.parse('2026-08-14T12:00:00Z')

@@ -22,10 +23,38 @@ test('selects the newest fresh aggregate-only regional rehome inventory', () =>
    targetRegistered: 1,
    completedLast24Hours: 9,
    abortedLast24Hours: 0,
+    hostNotArrivedLast24Hours: null,
    oldestActiveAgeMs: 30_000
  })
 })

+test('reads a line the director grew a field on, wherever the field sits', () => {
+  const result = parseRegionalRehomeInventory([{
+    timestamp: '2026-08-14T11:58:00Z',
+    textPayload:
+      '[orca-relay] regional rehome inventory hostNotArrivedLast24Hours=4 active=2' +
+      ' awaitingReceipt=1 targetRegistered=1 completedLast24Hours=9 abortedLast24Hours=7' +
+      ' oldestActiveAgeMs=30000 someFieldFromALaterRelease=11'
+  }], { now, maxAgeMs: 5 * 60_000 })
+  assert.equal(result.hostNotArrivedLast24Hours, 4)
+  assert.equal(result.abortedLast24Hours, 7)
+  assert.equal(result.oldestActiveAgeMs, 30_000)
+})
+
+test('reports an unmeasured host-not-arrived count as absent, not as zero', () => {
+  const [withField, withoutField] = ['4', null].map((value) =>
+    parseRegionalRehomeInventory([{
+      timestamp: '2026-08-14T11:58:00Z',
+      textPayload:
+        '[orca-relay] regional rehome inventory active=0 awaitingReceipt=0 targetRegistered=0' +
+        ' completedLast24Hours=0 abortedLast24Hours=0 oldestActiveAgeMs=none' +
+        (value === null ? '' : ` hostNotArrivedLast24Hours=${value}`)
+    }], { now, maxAgeMs: 5 * 60_000 })
+  )
+  assert.equal(withField.hostNotArrivedLast24Hours, 4)
+  assert.equal(withoutField.hostNotArrivedLast24Hours, null)
+})
+
 test('rejects stale, malformed, and identity-bearing lookalikes', () => {
  assert.throws(() => parseRegionalRehomeInventory([{
    timestamp: '2026-08-14T11:00:00Z',
@@ -36,3 +65,71 @@ test('rejects stale, malformed, and identity-bearing lookalikes', () => {
    textPayload: '[orca-relay] regional rehome inventory active=0 hostId=secret'
  }], { now }), /no aggregate/)
 })
+
+test('keeps out an identity-bearing field riding along on a complete line', () => {
+  const complete =
+    '[orca-relay] regional rehome inventory active=0 awaitingReceipt=0 targetRegistered=0' +
+    ' completedLast24Hours=0 abortedLast24Hours=0 oldestActiveAgeMs=none'
+  for (const extra of [' hostId=secret', ' userId=someone@example.test', ' note=a b']) {
+    assert.throws(
+      () => parseRegionalRehomeInventory(
+        [{ timestamp: '2026-08-14T11:59:00Z', textPayload: complete + extra }],
+        { now }
+      ),
+      /no aggregate/,
+      extra
+    )
+  }
+})
+
+test('refuses a line missing a required field, or repeating one', () => {
+  const missing =
+    '[orca-relay] regional rehome inventory active=0 awaitingReceipt=0 targetRegistered=0' +
+    ' completedLast24Hours=0 oldestActiveAgeMs=none'
+  assert.throws(
+    () => parseRegionalRehomeInventory(
+      [{ timestamp: '2026-08-14T11:59:00Z', textPayload: missing }],
+      { now }
+    ),
+    /no aggregate/
+  )
+  assert.throws(
+    () => parseRegionalRehomeInventory(
+      [{ timestamp: '2026-08-14T11:59:00Z', textPayload: `${missing} abortedLast24Hours=0 abortedLast24Hours=1` }],
+      { now }
+    ),
+    /no aggregate/
+  )
+  assert.throws(
+    () => parseRegionalRehomeInventory(
+      [{
+        timestamp: '2026-08-14T11:59:00Z',
+        textPayload: missing.replace('active=0', 'active=none') + ' abortedLast24Hours=0'
+      }],
+      { now }
+    ),
+    /no aggregate/
+  )
+})
+
+// The third edge of the chain the enable workflow depends on. The formatter is
+// pinned against this parser in the relay package's inventory-line census; this
+// pins the parser against the summary an operator reads, so a field that
+// reaches the evidence JSON and stops there fails here.
+test('publishes every parsed counter in the operator step summary', () => {
+  const job = readRelayWorkflow('operate-relay-production-rehome-job.yml')
+  // The jq program and the file it reads sit on separate continuation lines, so
+  // match the whole render rather than one line of it.
+  const summary = /jq -r '([^']*)' \\\n\s*"\$\{RUNNER_TEMP\}\/relay-rehome-inventory\.json"/.exec(job)?.[1]
+  assert.ok(summary, 'the rehome job no longer renders the inventory evidence')
+  const evidence = parseRegionalRehomeInventory([{
+    timestamp: '2026-08-14T11:58:00Z',
+    textPayload:
+      '[orca-relay] regional rehome inventory active=0 awaitingReceipt=0 targetRegistered=0' +
+      ' completedLast24Hours=0 abortedLast24Hours=0 hostNotArrivedLast24Hours=0 oldestActiveAgeMs=none'
+  }], { now, maxAgeMs: 5 * 60_000 })
+  for (const key of Object.keys(evidence)) {
+    if (key === 'timestamp') continue
+    assert.ok(summary.includes(`.${key}`), `${key} is missing from the step summary`)
+  }
+})