Verify hero_proc + hero_router + hero_codescalers start together cleanly #12
Loading…
Add table
Add a link
Reference in a new issue
No description provided.
Delete branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Parent: #8
Context
These three services must be running together as the baseline stack. Currently there may be socket mismatches or registration issues.
What to verify
hero_procstarts and its socket is up:$HERO_SOCKET_DIR/hero_proc/rpc.sockhero_routerstarts and maps TCP → correct Unix socketshero_codescalers_serverregisters under the correct socket:$HERO_SOCKET_DIR/hero_codescalers_server<N>/rpc.sockhero_codescalers_uiis reachable via hero_routerRelevant paths
hero_codescalers/crates/hero_codescalers_server/— RPC daemonhero_codescalers/crates/hero_codescalers_ui/— web dashboardhero_router)Acceptance Criteria
Verification on kristof4 — all acceptance criteria pass
Stack: hero_proc → hero_router → hero_codescalers (instance 0), running as user
mahmoud.Pre-condition fix
The pre-existing build on kristof4 was at commit
84dc730(still iroh-KVS, requiredHERO_CODESCALERS_KVS_NAMESPACE_SECRET). Pulled toc0c5957(sled-backed local store, no secret needed) viaservice_codescalers install --update --resetand restarted. After that the server boots cleanly:Acceptance criteria
hero_proc starts and its socket is up.
hero_router starts and maps TCP → correct Unix sockets.
hero_codescalers_server registers under the correct socket.
hero_codescalers_ui is reachable via hero_router.
hero_router dashboard lists codescalers as a clickable service.
hero_proc dashboard shows the registered service.
Direct health check via the binary.
Notes
SECRET_CODESCALERSis required; with the sled refactor (c0c5957) it's no longer used. Worth updating the README in a follow-up./hero_codescalers_server/ui/(matches the socket dir namehero_codescalers_server), not/hero_codescalers/ui/.Registered self node: kristof4confirms the per-server self-registration that the meeting brief described — each kristof box will show its own state, and admins reach each box independently via that server's hero_router.Next
Moving to #9 — verify nu Action → Job pipeline by triggering a real codescalers feature and confirming the job lands in hero_proc with correct tags + nu interpreter.
Closed — verified
The three-process stack boots cleanly on kristof4 in this order:
service_proc start --root --reset— supervisor on/root/hero/var/sockets/hero_proc/rpc.sock.service_router start --root --reset— TCP entry point bound to[<host-TUN>]:9988, scans for sibling sockets and exposes them under/<service>/{ui,rpc}.service_codescalers start --root— registershero_codescalers_serverandhero_codescalers_uias hero_proc-managed services; UI socket gets picked up by the router on its next scan and surfaces at/hero_codescalers_server/ui/.All three run as root, share
HERO_SOCKET_DIR=/root/hero/var/sockets, and survive supervisor restart cleanly. Verified thathero_codescalersis per-server admin (PR #140 enforces--rootand refuses per-user start).Stack is now the canonical setup for everything in #8.