A missing first article is an inspection result, not a failure

Eweka answers 'no such article' after ~450 ms; a quarter of inspection
requests were for missing articles and each failed job was retried up
to three times. Files get content_type 'missing' instead; homelab also
records it from failure reports of nodes not rebuilt yet.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

app/services/ContentService.ts

@@ -156,6 +156,11 @@ class ContentService {
     const binary = await db.from('binaries').where('file_id', task.fileId).first()
     const { contentType } = probe
     await db.from('files').where('id', file.id).update({ content_type: contentType })
+    if (contentType === 'missing') {
+      // Nothing to read; the binary needn't wait for inspection any more.
+      if (binary) await db.from('binaries').where('id', binary.id).update({ awaiting_inspection: null })
+      return result
+    }
 
     // Fingerprint, so par2 sets (now or later) can name this file exactly.
     if (binary) {
@@ -268,6 +273,7 @@ export function describeContent(filename: string, content: ContentResult | null)
     )
   }
   if (content?.named) lines.push(`"${filename}" got its real name from par2.`)
+  if (content?.contentType === 'missing') lines.push(`"${filename}": the server no longer has its first article.`)
   if (content?.contentType === 'nzbe') lines.push(`"${filename}" is an encrypted NZB container.`)
   if (content?.contentType === 'rclone') lines.push(`"${filename}" is rclone-encrypted.`)
   if (content?.yencName) lines.push(`"${filename}" is "${content.yencName}" (yEnc name).`)

app/services/RemoteContentService.ts

@@ -7,7 +7,7 @@ import QueueService from '#services/QueueService'
 import NntpService from '#services/NntpService'
 import ContentService, { describeContent } from '#services/ContentService'
 import ArchiveWalkService from '#services/ArchiveWalkService'
-import type { ContentTask, InspectProbe, InspectTask, WalkProbe } from '#lib/content/probe'
+import { isMissingArticle, type ContentTask, type InspectProbe, type InspectTask, type WalkProbe } from '#lib/content/probe'
 
 /** Inspection first; archive walks (D40) only when it has nothing waiting. */
 const QUEUES = ['collection-queue', 'archive-walk-queue'] as const
@@ -90,6 +90,11 @@ class RemoteContentService {
   /** The probe failed on the node (e.g. a missing article): retried like a local failure. */
   async fail(queue: QueueName, jobId: string, token: string, error: string) {
     const job = await this.job(queue, jobId)
+    // A missing first article from a node that predates the probe's own handling: a result,
+    // not something to retry (each try costs Eweka ~450 ms; LOG.md 2026-10-06 (14)).
+    if (queue === 'collection-queue' && job.data.remote && isMissingArticle(error)) {
+      return this.finish(queue, jobId, token, { size: 0, contentType: 'missing', md5_16k: null, yencName: '' })
+    }
     await this.locked(() => job.moveToFailed(new Error(`node: ${error}`), token, false))
   }
 

lib/content/probe.ts

@@ -83,6 +83,11 @@ export function chunkMd5(data: Buffer, fileSize: number): string | undefined {
   return createHash('md5').update(data.subarray(0, FIRST_CHUNK)).digest('hex')
 }
 
+/** A "no such article" answer: 430 from the news server, 410 from the seed relay. */
+export function isMissingArticle(error: unknown): boolean {
+  return /(^|[^0-9])(430|410)([^0-9]|$)/.test(String((error as any)?.message ?? error)) && !/timed out|no answer/i.test(String((error as any)?.message))
+}
+
 /** Decodes a body; some tools declare a size that leaves out a header (nzbe), the CRC still has to match. */
 export function decodeBody(body: Buffer): DecodedSegment {
   try {
@@ -132,7 +137,17 @@ export async function probeInspection(
   fetch: BodyFetcher,
   warn: (message: string) => void = () => {}
 ): Promise<InspectProbe> {
-  const segment = decodeBody(await fetch(task.segments[0].id))
+  let body: Buffer
+  try {
+    body = await fetch(task.segments[0].id)
+  } catch (error: any) {
+    // The server no longer has it (430; 410 through the seed relay), typically a takedown. A final
+    // answer, and an expensive one: Eweka takes ~450 ms per missing article, so it isn't retried
+    // (LOG.md 2026-10-06 (14)).
+    if (isMissingArticle(error)) return { size: 0, contentType: 'missing', md5_16k: null, yencName: '' }
+    throw error
+  }
+  const segment = decodeBody(body)
   const size = segment.meta.begin.size
   const contentType = sniffContentType(segment.data)
   const probe: InspectProbe = {

lib/content/sniff.ts

@@ -6,6 +6,8 @@
  * (LOG.md 2026-09-30 (3)): a fixed header, then data with 8 bits/byte of entropy.
  */
 export type ContentType =
+  /** Not sniffed: the server no longer has the file's first article (inspection records it, no retries). */
+  | 'missing'
   | 'nzbe'
   | 'par2'
   | 'rar4'

tests/functional/seed_api.spec.ts

@@ -234,11 +234,28 @@ test.group('Seed API', (group) => {
     ;(await client.post('/api/v1/seed/inspect/release').json({ queue: first.queue, jobId: first.jobId, token: first.token }).headers(auth)).assertStatus(200)
     const again = await lease()
     assert.equal(again.jobId, first.jobId)
-    ;(await client.post('/api/v1/seed/inspect/failed').json({ queue: again.queue, jobId: again.jobId, token: again.token, error: '430 no such article' }).headers(auth)).assertStatus(200)
+    ;(await client.post('/api/v1/seed/inspect/failed').json({ queue: again.queue, jobId: again.jobId, token: again.token, error: 'NNTP BODY: no answer within 120 s' }).headers(auth)).assertStatus(200)
     // Retried later (backoff), like a local failure.
     assert.include(['delayed', 'waiting'], await (await Job.fromId(QueueService.collectionQueue, again.jobId))!.getState())
   })
 
+  test('a missing first article reported by a node is recorded, not retried', async ({ client, assert }) => {
+    await QueueService.collectionQueue.drain()
+    await QueueService.archiveWalkQueue.drain()
+    const token = await SeedOrchestrator.issueToken('inspect-node-3')
+    const auth = { authorization: `Bearer ${token}` }
+    const now = new Date()
+    const [file] = await FileStore.insert([
+      { filename: 'Wq7n3LpX9vT2rK8mZ5cYbA.bin', file_key: 'inspect-node-file-3'.padEnd(40, '0'), poster: 'x', date: now.getTime(), parts: 1, groups: '[]', segments: [{ number: 1, id: 'gone@test', size: 1 }], created_at: now, updated_at: now },
+    ])
+    await QueueService.collectionQueue.add('process-collection', { fileId: file.id }, { jobId: `node-test-${file.id}-${Date.now()}` })
+    const job = (await client.post('/api/v1/seed/inspect/lease').json({ max: 1 }).headers(auth)).body().jobs[0]
+    const error = 'Seed API /api/v1/seed/articles?id=gone%40test: 410 {"error":"No such article"}'
+    ;(await client.post('/api/v1/seed/inspect/failed').json({ queue: job.queue, jobId: job.jobId, token: job.token, error }).headers(auth)).assertStatus(200)
+    assert.equal((await db.from('files').where('id', file.id).first()).content_type, 'missing')
+    assert.equal(await (await Job.fromId(QueueService.collectionQueue, job.jobId))!.getState(), 'completed')
+  })
+
   test('a range reported to its end but never finished is marked done, not leased', async ({ client, assert }) => {
     const token = await SeedOrchestrator.issueToken('late-node')
     const [range] = await db.table('seed_ranges').insert({ group_name: 'alt.binaries.test', start_id: 1, end_id: 100, next_id: 101 }).returning('id')

tests/unit/probe.spec.ts

@@ -71,9 +71,16 @@ test.group('content probe', () => {
     assert.lengthOf(probe.archive!.entries, 0)
   })
 
-  test('a missing article fails the probe (the job is retried)', async ({ assert }) => {
+  test('a missing first article is a result, not a failure (no retries)', async ({ assert }) => {
     const { task, fetch } = posted('gone.bin', Buffer.alloc(10, 1))
     task.segments[0].id = 'missing@test'
-    await assert.rejects(() => probeInspection(task, fetch), /430/)
+    assert.include(await probeInspection(task, fetch), { contentType: 'missing' })
+    // Through the seed relay: 410.
+    const relayed = await probeInspection(task, async () => {
+      throw new Error('Seed API /api/v1/seed/articles?id=x: 410 {"error":"No such article"}')
+    })
+    assert.equal(relayed.contentType, 'missing')
+    // Anything else still fails (and is retried).
+    await assert.rejects(() => probeInspection(task, async () => { throw new Error('NNTP BODY: no answer within 120 s') }), /no answer/)
   })
 })