Expand description
The renewal scheduler: one task per managed certificate that decides when
to (re)issue, takes the storage lease so only one instance orders, runs
order::issue, persists, publishes into ManagedCerts, and backs off on
failure. Instances that lose the lease poll storage and adopt whatever the
leaseholder wrote, refreshing challenge certs meanwhile so the CA may
validate through any of them.
Structs§
- Abort
OnDrop 🔒 - Aborts the wrapped task on drop — including when the future holding it is
simply cancelled (e.g. a slot task dropped on shutdown) rather than run to
completion, so a lease-keepalive loop can never outlive the order it was
keeping alive for. A plain
abort()call reached only via a specific success/error path does not cover that case. - Managed
Slot - Manager
- Manager
Config - Slot
Control 🔒
Enums§
Constants§
- ARI_
INTERVAL_ 🔒SECS - ARI is re-queried at most this often per certificate.
- LEASE_
TTL - PLACEHOLDER_
REMINT_ 🔒MARGIN_ SECS - Placeholders are minted with one hour of validity; re-mint this long before
they lapse so a slot stuck in
Placeholdernever serves an expired certificate (design section 4: “re-minted on every restart and hourly while still in Placeholder state”).
Functions§
- backoff_
secs 60·2^(failures−1)seconds, capped at one hour;0when there are no failures.- when_
to_ renew - Unix time at which the certificate should be renewed:
renew_beforeahead of expiry, or the start of the CA’s ARI window if that comes first.