Skip to content

docs(sensitive): Paperless-ngx's acceptance on trinity (#133) - #700

Merged
Gerrrt merged 1 commit into
mainfrom
gerrrt/paperless-ngx-sensitive-14d2cb
Sep 28, 2026
Merged

Gerrrt merged 1 commit into
mainfrom
gerrrt/paperless-ngx-sensitive-14d2cb

Conversation

@Gerrrt

@Gerrrt Gerrrt commented Sep 28, 2026

Copy link
Copy Markdown
Owner

Records the acceptance #133's 2026-09-19 correction asked for: cpus: 4 and mem_limit: 3072m measured on trinity under a scan, not just stated.

What was measured

A synthetic backlog, chosen over real paperwork so nothing personal went through the archive: five one-page scans and one 50-page scan, 300 dpi, image-only and noised. Cgroup and latency samples were taken every 1–2 s.

memory.peak 1716 MiB of 3072m (56 %), no oom_kill
CPU 3.0 cores at the busiest minute; 9 throttled periods, 0.2 s in total
Time ~16 s per single page; 3 min 54 s for the 50 pages
Rest of the tier nothing unhealthy; Vaultwarden through Caddy ≤ 19 ms, no non-200s

Both limits stand. THREADS_PER_WORKER=3 does the limiting, and cpus: 4 is only the backstop. Memory is kept, not cut, on one sample. The one-month container_memory_rss re-derivation still applies.

Things worth knowing

  • A concurrent STACK=sensitive make backup from another session interrupted the first run 90 seconds into the 50 pages. The file stayed in consume/ and was consumed from scratch after the restart, so the numbers come from fresh cgroup counters. That backup, 20260928T203415Z, holds the five one-page test documents in its paperless-media archive. They are synthetic text and harmless, but that archive's library isn't empty.
  • The test documents were deleted. A hard delete from the Django shell removes the rows but leaves the files, so the 18 files under media/ were removed by hand and the index rebuilt. document_sanity_checker reports no issues, the library holds 0 documents and consume/ is empty.
  • make validate on trinity fails one check unrelated to this change: backup-offsite.sh --self-test, case "a DEST under /tmp is refused". /tmp on trinity is itself a tmpfs, so the tmpfs refusal fires first. Everything else passes.

This changes docs and comments only (docs/changelog.md, stacks/sensitive/README.md, stacks/sensitive/compose.yaml) and needs no redeploy.

Refs #133

🤖 Generated with Claude Code

Measured under a synthetic backlog (five one-page scans and one of 50
pages, 300 dpi, image-only): memory.peak 1716 MiB of 3072m with no OOM
kill, 3.0 cores at the busiest minute and 0.2 s throttled, 3 min 54 s for
the 50 pages, and Vaultwarden through Caddy never slower than 19 ms.
Both limits stand. The test documents were deleted afterwards.

Refs #133

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@Gerrrt
Gerrrt force-pushed the gerrrt/paperless-ngx-sensitive-14d2cb branch from f3755c2 to de75516 Compare September 28, 2026 20:57
@Gerrrt
Gerrrt merged commit ee17801 into main Sep 28, 2026
7 checks passed
@Gerrrt
Gerrrt deleted the gerrrt/paperless-ngx-sensitive-14d2cb branch September 28, 2026 21:02
Gerrrt added a commit that referenced this pull request Sep 28, 2026
…is deleted (#133) (#702)

20260928T203415Z held the five one-page acceptance documents. It is removed
from trinity and oracle; 20260928T204939Z, taken after the cleanup, was
checked identical on both sides first. Corrects the #700 entry.

Refs #133

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant