Remaining: add systemd sd_notify integration to archied - emit READY=1 once the run loop is up, and sample loop progress to WATCHDOG=1 so a stalled event loop trips the unit watchdog. Nothing else from the original gap list is outstanding.
Already shipped (do not rebuild): multi-check readiness returning JSON at /health/detailed (state_db, config parse, disk<90%, model, gateway contract ping) in internal/domain/health + internal/infrastructure/readiness; marker-file drain control with a boot_id + PID1 instantiation epoch in internal/infrastructure/drain; and an out-of-loop shutdown-leash watchdog in internal/app/archied/shutdown_watchdog.go.
Corrections to the earlier text: the drain marker is ${XDG_CONFIG_HOME:-~/.config}/archie/.drain_request.json, not ~/.archie/...; /health and /healthz are plain-text liveness and the JSON report is /health/detailed; the gateway probe is a contract ping, not a running/draining plus connected-count report. Reworded 2026-09-22 from an acceptance check; the original text is in Dolt history.
Notes
docs/prds/hermes-gap-analysis.md (referenced above) was intentionally deleted in commit 65aa5b4 ('chore: cleanup', 2026-07-28) after it had already spawned this and ~14 sibling beads issues from its 60 identified gaps — it is not a broken link to restore. This issue's description is the surviving record of its slice of that analysis.
Maintainer 2026-09-22: REWORD AND RE-CHECK, do not close or accept as-is. Evidence: drain control exists (internal/app/archied/drain.go) plus a shutdown watchdog, so part of the ask is shipped; the multi-check readiness shape (state_db/config/disk/model/gateway) is unverified. Next: verify each acceptance criterion against the tree, then reword the description to what actually remains.
Acceptance check 2026-09-22 (read-only, file:line): probes SHIPPED at infrastructure/readiness/probes.go:40 (store, read-only StatusCounts), :113 (config -> configuration.Validate), :169 (disk threshold 0.90), wired in app/archied/readiness.go:38-41; overall registry domain/health/health.go:96; drain marker infrastructure/drain/drain.go:23 with epoch at :26-72 and consumer :130, path at app/archied/drain.go:28-36; shutdown watchdog app/archied/shutdown_watchdog.go:38-105. sd_notify ABSENT: no NOTIFY_SOCKET or notify library anywhere in internal/, cmd/ or go.mod. Partial: the gateway criterion is a ContractProbe pinging Snapshot (readiness.go:45-48), not a draining/connected-count report.
Close reason
Landed on main (2eeef60b, re-gated as ecf42c8c). READY=1 is emitted at markServing - this tree single definition of serving - and WATCHDOG=1 beats are licensed by daemon.LastPollAt rather than a bare ticker, so a stalled loop goes silent and trips the unit watchdog; all of it no-ops without NOTIFY_SOCKET. Two follow-ups are filed separately: every shipped unit is Type=simple with no WatchdogSec, so the feature is inert until the archied unit changes, and archie-gateway/archie-state-store emit no READY.
Imported from beads archie-core-1786637496703-262-1e49aef1 · GitHub #283 · type feature · priority P2 · created 2026-08-08 · closed 2026-09-21
Remaining: add systemd sd_notify integration to archied - emit READY=1 once the run loop is up, and sample loop progress to WATCHDOG=1 so a stalled event loop trips the unit watchdog. Nothing else from the original gap list is outstanding.
Already shipped (do not rebuild): multi-check readiness returning JSON at /health/detailed (state_db, config parse, disk<90%, model, gateway contract ping) in internal/domain/health + internal/infrastructure/readiness; marker-file drain control with a boot_id + PID1 instantiation epoch in internal/infrastructure/drain; and an out-of-loop shutdown-leash watchdog in internal/app/archied/shutdown_watchdog.go.
Corrections to the earlier text: the drain marker is ${XDG_CONFIG_HOME:-~/.config}/archie/.drain_request.json, not ~/.archie/...; /health and /healthz are plain-text liveness and the JSON report is /health/detailed; the gateway probe is a contract ping, not a running/draining plus connected-count report. Reworded 2026-09-22 from an acceptance check; the original text is in Dolt history.
### Notes
docs/prds/hermes-gap-analysis.md (referenced above) was intentionally deleted in commit 65aa5b4 ('chore: cleanup', 2026-07-28) after it had already spawned this and ~14 sibling beads issues from its 60 identified gaps — it is not a broken link to restore. This issue's description is the surviving record of its slice of that analysis.
Maintainer 2026-09-22: REWORD AND RE-CHECK, do not close or accept as-is. Evidence: drain control exists (internal/app/archied/drain.go) plus a shutdown watchdog, so part of the ask is shipped; the multi-check readiness shape (state_db/config/disk/model/gateway) is unverified. Next: verify each acceptance criterion against the tree, then reword the description to what actually remains.
Acceptance check 2026-09-22 (read-only, file:line): probes SHIPPED at infrastructure/readiness/probes.go:40 (store, read-only StatusCounts), :113 (config -> configuration.Validate), :169 (disk threshold 0.90), wired in app/archied/readiness.go:38-41; overall registry domain/health/health.go:96; drain marker infrastructure/drain/drain.go:23 with epoch at :26-72 and consumer :130, path at app/archied/drain.go:28-36; shutdown watchdog app/archied/shutdown_watchdog.go:38-105. sd_notify ABSENT: no NOTIFY_SOCKET or notify library anywhere in internal/, cmd/ or go.mod. Partial: the gateway criterion is a ContractProbe pinging Snapshot (readiness.go:45-48), not a draining/connected-count report.
### Close reason
Landed on main (2eeef60b, re-gated as ecf42c8c). READY=1 is emitted at markServing - this tree single definition of serving - and WATCHDOG=1 beats are licensed by daemon.LastPollAt rather than a bare ticker, so a stalled loop goes silent and trips the unit watchdog; all of it no-ops without NOTIFY_SOCKET. Two follow-ups are filed separately: every shipped unit is Type=simple with no WatchdogSec, so the feature is inert until the archied unit changes, and archie-gateway/archie-state-store emit no READY.
---
_Imported from beads `archie-core-1786637496703-262-1e49aef1` · GitHub #283 · type feature · priority P2 · created 2026-08-08 · closed 2026-09-21_
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
Remaining: add systemd sd_notify integration to archied - emit READY=1 once the run loop is up, and sample loop progress to WATCHDOG=1 so a stalled event loop trips the unit watchdog. Nothing else from the original gap list is outstanding.
Already shipped (do not rebuild): multi-check readiness returning JSON at /health/detailed (state_db, config parse, disk<90%, model, gateway contract ping) in internal/domain/health + internal/infrastructure/readiness; marker-file drain control with a boot_id + PID1 instantiation epoch in internal/infrastructure/drain; and an out-of-loop shutdown-leash watchdog in internal/app/archied/shutdown_watchdog.go.
Corrections to the earlier text: the drain marker is ${XDG_CONFIG_HOME:-~/.config}/archie/.drain_request.json, not ~/.archie/...; /health and /healthz are plain-text liveness and the JSON report is /health/detailed; the gateway probe is a contract ping, not a running/draining plus connected-count report. Reworded 2026-09-22 from an acceptance check; the original text is in Dolt history.
Notes
docs/prds/hermes-gap-analysis.md (referenced above) was intentionally deleted in commit
65aa5b4('chore: cleanup', 2026-07-28) after it had already spawned this and ~14 sibling beads issues from its 60 identified gaps — it is not a broken link to restore. This issue's description is the surviving record of its slice of that analysis.Maintainer 2026-09-22: REWORD AND RE-CHECK, do not close or accept as-is. Evidence: drain control exists (internal/app/archied/drain.go) plus a shutdown watchdog, so part of the ask is shipped; the multi-check readiness shape (state_db/config/disk/model/gateway) is unverified. Next: verify each acceptance criterion against the tree, then reword the description to what actually remains.
Acceptance check 2026-09-22 (read-only, file:line): probes SHIPPED at infrastructure/readiness/probes.go:40 (store, read-only StatusCounts), :113 (config -> configuration.Validate), :169 (disk threshold 0.90), wired in app/archied/readiness.go:38-41; overall registry domain/health/health.go:96; drain marker infrastructure/drain/drain.go:23 with epoch at :26-72 and consumer :130, path at app/archied/drain.go:28-36; shutdown watchdog app/archied/shutdown_watchdog.go:38-105. sd_notify ABSENT: no NOTIFY_SOCKET or notify library anywhere in internal/, cmd/ or go.mod. Partial: the gateway criterion is a ContractProbe pinging Snapshot (readiness.go:45-48), not a draining/connected-count report.
Close reason
Landed on main (
2eeef60b, re-gated asecf42c8c). READY=1 is emitted at markServing - this tree single definition of serving - and WATCHDOG=1 beats are licensed by daemon.LastPollAt rather than a bare ticker, so a stalled loop goes silent and trips the unit watchdog; all of it no-ops without NOTIFY_SOCKET. Two follow-ups are filed separately: every shipped unit is Type=simple with no WatchdogSec, so the feature is inert until the archied unit changes, and archie-gateway/archie-state-store emit no READY.Imported from beads
archie-core-1786637496703-262-1e49aef1· GitHub #283 · type feature · priority P2 · created 2026-08-08 · closed 2026-09-21