Expand description
Re-adoption on daemon start (ARCHITECTURE.md §3: upgrade = restart +
re-adopt). Routes and supervision records live only in the daemon’s
memory, so they die with the daemon — re-adoption rebuilds them from
the checkpoint journal, the one durable truth. For every ACTIVE
instance, each start:* checkpoint whose recorded process is still
alive has its proxy routes and supervision record re-registered; a
dead process is noted, not restarted (v0 supervision policy: observe,
don’t restart).
Structs§
- Adoption
Summary - What one adoption pass observed — for the daemon log only.
Functions§
- readopt
- Rebuild routing and supervision from the journal. Opens the store fresh (short-lived access; the store is multi-process-safe). Errors reading the store are non-fatal: re-adoption is best-effort recovery, and the daemon must come up regardless.