HydraCache 0.57.0
HydraCache 0.57.0
0.57.0 adds a read-only Management Center and completes the observability
surface around the already-shipped server, actuator, grid, and operator pieces.
The release replaces modeled-only admin status with an explicit live/modeled
source tag, serves Prometheus and overview data from the internal admin surface,
ships a same-origin console, and adds a drift-guarded Grafana dashboard.
The release is an operate-in-prod surface, not a new control plane. Mutations
remain on the 0.56 authz-gated admin API and operator paths.
Highlights
- Added
ClusterStatusProviderwithsource:"live"|"modeled"so modeled data
is never presented as live. - Completed Prometheus rendering for admission, topology, and cluster-grid
metrics already registered by the observability crate. - Served
/metrics,/cluster/overview,/console, and the read-only actuator
from the internal admin surface, including during drain. - Added the
ClusterOverviewread-model with honest leader, reachability,
consistency distribution, partition, and backup-age semantics. - Added the read-only Management Center web console and wired its Node-gated
tests intocargo xtask verifywhen Node/npm are available. - Wired
role = "member"to build a real in-process grid member for W6a while
leavinglocalandclientroles modeled. - Added
docs/observability/dashboards/hydracache-overview.jsonand a Rust
drift guard that validates dashboard PromQL againstregistered_metric_names().
Verification
Focused gates:
cargo test -p hydracache-server --locked cluster_status
cargo test -p hydracache-observability --locked prometheus_exporter
cargo test -p hydracache-server --locked admin_http
cargo test -p hydracache-server --locked cluster_overview
npm --prefix console test
cargo test -p hydracache-server --test grid_host --locked -j 1
cargo test -p hydracache-server --test actuator_mount --locked -j 1
cargo test -p hydracache-observability --test dashboard_metrics --locked -j 1Full gate:
cargo xtask verifyDeferred Technical Debt
W6b networked multi-daemon grid hosting is intentionally deferred and tracked as
TD-0008. The
daemon now hosts a real in-process member for source:"live" status, but it
does not yet wire the networked raft, transport, and discovery adapters into a
multi-process daemon cluster. The ignored sentinel test remains in place for
that follow-up.
Scope Boundary
Management Center is read-only and runs on the internal admin surface. It does
not add a write UI, bundled TSDB, new consistency level, or new fast-path cache
algorithm.
Release Status
Published as 0.57.0 after W0-W8 landed and the full cargo xtask verify gate
passed. The remaining networked daemon-grid follow-up is documented as
TD-0008.