Module cluster 

Available on crate feature serve only.

Clustered execution (#197, Mode A): when --cluster is set, every instance runs a claim loop that pulls Pending runs from the shared SQL history DB, so submissions pull-balance across instances and a crashed instance’s runs are re-run by a survivor. Inert unless enabled.

Structs

ClusterConfig

Validated cluster settings, derived from --cluster* args.

ClusterHandle

Cheaply-cloneable runtime handle for cluster coordination, held in ServerState. Carries the kick signal (so submit can wake the local claim loop immediately) and the cached live-member count (so /readyz need not hit the DB per probe).

Functions

claim_loop

Background claim + cancel-propagation loop (cluster mode only). Each wake: