Chill Bits

Technical Journal — Homework Archiving & Operational Reality

—

Date: Today
System: Alice
Theme: Archiving, storage truth, and learning how production really behaves

What I worked on

Today was about turning “archive homework” from an idea into a production-safe operation.

I built and iterated on an alice:homework:archive command with very strict constraints:

  • Dry-run by default
  • Reversible where possible
  • Explicit logging
  • No silent failure

The command does more than just flip a status:

  • Validates Google Drive destination
  • Copies homework files from R2 to Google Drive
  • Handles Shared Drive constraints
  • Cleans up R2 only after successful copy
  • Supports idempotency (already archived homework becomes cleanup-only)

Once the manual command was stable, I designed the automation layer:

  • A scheduled task at 03:00 AM
  • Finds homework older than 12 days
  • Dispatches archive jobs to the queue (no long work in scheduler)
  • All actions logged into a dedicated daily incident log:storage/logs/homework-archive-YYYY-mm-dd.log with structured, parseable entries for future reporting.

Problems I hit (and why they mattered)

This was not a “happy path” day.

  1. Google Drive + Service Accounts are unforgiving
    • Service accounts have no storage quota in My Drive.
    • Shared Drives are mandatory.
    • Even then, permissions behave differently than expected.
  2. Drive API lies politely
    • 404 File not found doesn’t always mean “not found”.
    • It can mean:
      • wrong drive context
      • missing permission
      • operation not allowed (delete) even when create is allowed
    • The API surface does not clearly distinguish these cases.
  3. Preflight checks can become self-sabotage
    • I added a “touch test” (create + delete temp file).
    • Creation succeeded.
    • Deletion failed with 404.
    • The system was actually writable — but my preflight logic said “fail”.
    This forced an important correction:Write success proves writability. Cleanup is optional.
  4. Shared Drive permissions are asymmetric
    • A service account may:
      • create files
      • but not delete or trash them
    • That is normal and acceptable.
    • The system must not treat that as a hard failure.

Key changes I made

  • Redefined preflight success:
    • Create success = writable
    • Delete/trash = best-effort, non-blocking
  • Improved diagnostics:
    • Clearly separated driveId vs parentFolderId
    • Logged exact IDs and operations
  • Ensured safety guarantees:
    • Never delete from R2 unless Drive copy succeeded
    • Never archive homework if copying is incomplete
  • Added operational structure:
    • Queue jobs for long-running work
    • Scheduled discovery only
    • Single, clean, parseable log per day for incidents

What I learned

  1. Infrastructure truth beats assumptions
    • APIs don’t behave how docs suggest in edge cases.
    • Production behavior is the only reliable source.
  2. Preflight checks must prove capability, not cleanliness
    • “Can I write?” matters.
    • “Can I clean up?” is secondary.
  3. Archiving is a lifecycle, not a flag
    • Status change
    • External storage migration
    • Source cleanup
    • Idempotency
    • Automation
      All of these must align or the system rots quietly.
  4. Good logs are future leverage
    • By logging every archive action into a dedicated daily file,
      I’ve set up:
      • future reports
      • incident reviews
      • confidence in automation
    • This is observability, not noise.

State at end of day

  • Manual archive command is operationally correct
  • Shared Drive quirks are understood and handled
  • R2 cleanup is safe and intentional
  • Auto-archive design is ready and disciplined
  • The system feels calmer, not more complex

Today wasn’t about speed.
It was about earning trust in a system that deletes data — and that’s the kind of work that actually matters.