Wire repo-recall into the ser8 o11y plane (SigNoz log parser + verify) #115

New issue

Closed

opened 2026-06-18 07:28:10 +00:00 by coilysiren · 0 comments

coilysiren commented

2026-06-18 07:28:10 +00:00

Owner

Goal

Bring repo-recall's error monitoring in line with every other kai-server service: a version-controlled SigNoz log parser so its logs land in the normalized schema and its errors surface in the same ser8 SigNoz pane as o2r/atlas/etc. This is the "error monitoring, in sync with everything else" half of the revival. (OTLP emission env is owned by the deploy-onboarding issue; this issue owns the parse + verify side.)

Context

The fleet o11y plane is SigNoz + ClickHouse on ser8 (coilyco-bridge/deploy services/ser8-observability/). Log parsing is version-controlled in services/signoz-pipelines/pipelines/*.json, applied by apply.py (SigNoz at http://ser8:30808, admin creds from SSM /coilysiren/signoz-ser8/admin-password), drift-checked by coily exec check-signoz-pipelines. The field-normalization contract is in services/ser8-observability/docs/log-schema.md; producer inventory in docs/log-producers.md.

Work

Study an existing services/signoz-pipelines/pipelines/<NN>-*.json parser and ser8-observability/docs/log-schema.md.
Add services/signoz-pipelines/pipelines/<NN>-repo-recall.json matching repo-recall's log format, coercing its fields into the normalized schema (level, service, message, etc.). Add repo-recall to the producer inventory in docs/log-producers.md.
Apply + verify: coily exec apply-signoz-pipelines then coily exec check-signoz-pipelines must be clean (no drift).
Confirm in the ser8 SigNoz that repo-recall logs appear and an error-level line is parsed into the normalized schema.

Acceptance

A repo-recall ingest pipeline exists in services/signoz-pipelines/pipelines/.
coily exec check-signoz-pipelines is clean after apply.
repo-recall logs are visible and correctly parsed in the ser8 SigNoz pane.
docs/log-producers.md lists repo-recall.

Why headless

All code: a JSON parser file, a docs-inventory line, an apply + a drift check. SigNoz admin creds resolve from SSM via the existing apply.py. No human gate, no design fork.

Coordinates with

The deploy-onboarding issue (it adds the OTLP emission env to services/repo-recall/deploy/main.yml). The log parser here works on the container logs regardless, so this can land independently; full trace/log richness arrives once emission lands.

## Goal Bring repo-recall's error monitoring in line with every other kai-server service: a version-controlled SigNoz log parser so its logs land in the normalized schema and its errors surface in the same ser8 SigNoz pane as o2r/atlas/etc. This is the "error monitoring, in sync with everything else" half of the revival. (OTLP *emission* env is owned by the deploy-onboarding issue; this issue owns the *parse + verify* side.) ## Context The fleet o11y plane is SigNoz + ClickHouse on ser8 (`coilyco-bridge/deploy` `services/ser8-observability/`). Log parsing is version-controlled in `services/signoz-pipelines/pipelines/*.json`, applied by `apply.py` (SigNoz at `http://ser8:30808`, admin creds from SSM `/coilysiren/signoz-ser8/admin-password`), drift-checked by `coily exec check-signoz-pipelines`. The field-normalization contract is in `services/ser8-observability/docs/log-schema.md`; producer inventory in `docs/log-producers.md`. ## Work 1. Study an existing `services/signoz-pipelines/pipelines/<NN>-*.json` parser and `ser8-observability/docs/log-schema.md`. 2. Add `services/signoz-pipelines/pipelines/<NN>-repo-recall.json` matching repo-recall's log format, coercing its fields into the normalized schema (level, service, message, etc.). Add repo-recall to the producer inventory in `docs/log-producers.md`. 3. Apply + verify: `coily exec apply-signoz-pipelines` then `coily exec check-signoz-pipelines` must be clean (no drift). 4. Confirm in the ser8 SigNoz that repo-recall logs appear and an error-level line is parsed into the normalized schema. ## Acceptance - A `repo-recall` ingest pipeline exists in `services/signoz-pipelines/pipelines/`. - `coily exec check-signoz-pipelines` is clean after apply. - repo-recall logs are visible and correctly parsed in the ser8 SigNoz pane. - `docs/log-producers.md` lists repo-recall. ## Why headless All code: a JSON parser file, a docs-inventory line, an apply + a drift check. SigNoz admin creds resolve from SSM via the existing apply.py. No human gate, no design fork. ## Coordinates with The deploy-onboarding issue (it adds the OTLP emission env to `services/repo-recall/deploy/main.yml`). The log parser here works on the container logs regardless, so this can land independently; full trace/log richness arrives once emission lands.