bot attempt to figure the bug, based on reproducible example

[vagkaratzas]

Related to issue #6884

I (and other people in my team) have long been suffering from seemingly random cache failures of Nextflow, here and there. I think I have tracked down the culprit and posted a reproducible example on the relevant issue above and at a blog post.

I fed this example to the bots and they tried to pinpoint the culprit, so this PR is to potentially give the direction to where the bug-fix might be hiding.

Hope this helps!

Bot report:

Root Cause

  The bug is a hash chain mismatch between the retry path and the resume path.

  checkCachedOrLaunchTask builds a chain:
  H1 = hash(H_base + 1)
  H2 = hash(H1   + 2)
  H3 = hash(H2   + 3)   ← tries increments each iteration

  The resume always starts with tries=1 from H_base (a fresh task with attempt=1), correctly traversing the chain.

  The retry (old code) called checkCachedOrLaunchTask(taskCopy, taskCopy.hash, false), where taskCopy.hash was the chained hash of the last
  failed attempt (e.g. H2) — not H_base. Then internally used tries = task.failCount + 1.

  In the user's scenario (one killed prior attempt + one OOM failure before the successful retry):

  ┌─────────────────┬───────────────┬────────────┬──────────────────────────┐
  │      Path       │ Starting hash │ tries used │          Result          │
  ├─────────────────┼───────────────┼────────────┼──────────────────────────┤
  │ Retry (old)     │ H2            │ 2          │ hash(H2 + 2) = H3_retry  │
  ├─────────────────┼───────────────┼────────────┼──────────────────────────┤
  │ Resume (step 3) │ H_base → ...  │ 3          │ hash(H2 + 3) = H3_resume │
  └─────────────────┴───────────────┴────────────┴──────────────────────────┘

  H3_retry ≠ H3_resume — the resume never finds the cached result.

  Fix (3 changes to TaskProcessor.groovy)

  1. checkCachedOrLaunchTask (line 799): tries = task.failCount + 1 → tries = 1. The loop always starts at position 1 and naturally advances
  past any existing directories.
  2. Path 1 retry (spot preemption / transient errors): Inside the submit block, set taskCopy.config.attempt = 1 and compute retryHash = new
  TaskHasher(taskCopy).compute() (= H_base). Pass retryHash instead of taskCopy.hash.
  3. Path 2 retry (RETRY error strategy): Same pattern — compute H_base with attempt=1 before taskCopy.config.attempt is bumped to
  taskErrCount+1. The loop then traverses H1→H2→H3 exactly as the resume will, landing on the same hash position.

[netlify]

✅ Deploy Preview for nextflow-docs-staging ready!

Name	Link
🔨 Latest commit	faf89cd
🔍 Latest deploy log	https://app.netlify.com/projects/nextflow-docs-staging/deploys/69a709f87b84a80009ad9fbd
😎 Deploy Preview	https://deploy-preview-6882--nextflow-docs-staging.netlify.app
📱 Preview on mobile	Toggle QR Code... Use your smartphone camera to open QR code link.

---

To edit notification comments on pull requests, go to your Netlify project configuration.