bug: backend image missing redis_breaker_util.py — Dockerfile COPY not added for new top-level module (#986)

[vybe]

Summary

Commit 0c671158 (feat: per-agent dispatch circuit breaker (#526) (#986)) adds a new top-level backend module src/backend/redis_breaker_util.py, imported by services/agent_client.py, but does not add a corresponding COPY line to docker/backend/Dockerfile. The backend Dockerfile copies top-level modules by explicit name, so the new file never lands in the image. The backend then crash-loops on startup with ModuleNotFoundError, taking down the whole platform (frontend/mcp-server/scheduler are gated on backend health via depends_on). Rebuilding with --no-cache does not help — the file is simply never copied.

Component

Backend / build packaging (docker/backend/Dockerfile)

Priority

P0 — backend fails to start; full platform outage on any deploy of this commit.

Error

ModuleNotFoundError: No module named 'redis_breaker_util'
  File "/app/main.py", line 44, in <module>
    from routers.auth import router as auth_router
  File "/app/routers/agents.py", line 41, in <module>
    from services.agent_service import (...)
  File "/app/services/agent_service/lifecycle.py", line 27, in <module>
    from services.skill_service import skill_service
  File "/app/services/skill_service.py", line 26, in <module>
    from services.agent_client import get_agent_client, AgentClientError
  File "/app/services/agent_client.py", line 31, in <module>
    from redis_breaker_util import (...)

Location

• File: docker/backend/Dockerfile (top-level module COPY block, ~lines 62–68)
• Missing module: src/backend/redis_breaker_util.py
• Importer: src/backend/services/agent_client.py:31

Root Cause

docker/backend/Dockerfile enumerates top-level backend modules individually:

COPY ../../src/backend/main.py /app/
COPY ../../src/backend/config.py /app/
COPY ../../src/backend/models.py /app/
COPY ../../src/backend/db_models.py /app/
COPY ../../src/backend/dependencies.py /app/
COPY ../../src/backend/database.py /app/
COPY ../../src/backend/logging_config.py /app/

Subdirectories (routers/, services/, adapters/, utils/, db/, canary/) are copied wholesale, so the other new files in #986 (e.g. services/dispatch_breaker.py, services/circuit_breaker_view.py) are present — but the new top-level redis_breaker_util.py is not in the explicit list and is silently dropped. The module is present in source; it just never reaches /app in the image.

DB migrations from this commit (idempotency_keys table, schedule_executions.loop_id) run successfully before the import crash, confirming the DB layer is fine — this is purely a packaging defect.

Reproduction Steps

1. Check out 0c671158 (or current dev).
2. Build the backend image: docker compose -f docker-compose.prod.yml build backend (with or without --no-cache).
3. Start the stack: docker compose -f docker-compose.prod.yml up -d.
4. docker logs trinity-backend → ModuleNotFoundError: No module named 'redis_breaker_util'; backend never becomes healthy; dependent services stay in Created.

Suggested Fix

Add the missing COPY for the new module:

# docker/backend/Dockerfile, alongside the other top-level modules
COPY ../../src/backend/redis_breaker_util.py /app/

Stronger fix — eliminate this entire class of bug by copying top-level modules with a glob so future additions don't silently drop out:

COPY ../../src/backend/*.py /app/

Recommended guardrail: add a build-time import smoke check to the backend image build (e.g. RUN python -c "import main"), so a missing module fails the build instead of shipping a broken image.

Environment

• Trinity version: 0c671158 (dev, v0.6.0 + reliability commits)
• Affected since: feat: per-agent dispatch circuit breaker to prevent backlog poisoning (#526) #986 merge
• Deploy: docker-compose.prod.yml, non-root backend image

Related

• docker/backend/Dockerfile
• src/backend/services/agent_client.py
• PR feat: per-agent dispatch circuit breaker to prevent backlog poisoning (#526) #986 / feat: per-agent dispatch circuit breaker to prevent backlog poisoning (RELIABILITY-007) #526 (per-agent dispatch circuit breaker)

[alucard210319]

Hi, I can take a focused pass at this backend image packaging issue.

The immediate fix looks small, but I would also add a narrow verification guard so future top-level backend modules do not silently get dropped from the image. I would keep the PR scoped to the Dockerfile copy behavior plus a build/import smoke check if it fits your current test setup.

If outside paid help is appropriate, I can do a small verified PR for $35. If you only want the root-cause/patch plan documented, I can do that for $15.

Recent example: I helped LiveReview fix stale Docker build images, and the maintainer merged the fix through their internal PR.

[vybe]

Fixed by #1034 (squash-merged to dev as 65515ff). Dockerfile COPY collapsed to a *.py glob so new top-level modules can't be dropped from the prod image; backend-image-smoke.yml CI now guards the baked prod image against this class of defect. Closing manually — the PR body's wording didn't trigger GitHub's auto-close.