HydraCache 0.41.0

HydraCache 0.41.0

0.41.0 defines the distributed cache grid roadmap and ships the first safe
implementation slice. This is intentionally not a full production data-grid
claim: the release adds correctness primitives, release gates, and diagnostics
needed before future grid hardening.

Highlights

• Finalized five cluster ADRs and linked them from docs/cluster/readiness.md.
• Replaced direct raft MemStorage coupling with a RaftLogStore seam and
  InMemoryRaftLogStore.
• Added deterministic primary + backup placement through
  ClusterReplicationStrategy, Replicas, EffectiveReplicationMap, and
  ReplicationConfig.
• Added rebalance plan-as-data primitives: RebalancePlan, RebalanceTask, and
  RebalanceTaskAck.
• Added versioned ReplicatedSlot tombstones with tombstone-wins-on-tie merge
  semantics, TombstoneBudget, and TombstoneTracker.
• Added opt-in value-replication builder knobs with mandatory byte-cap
  validation for client/member startup.
• Added replicated-value security posture:
  Replication::LocalOnly, ReplicationKeyProvider, redaction hook, and loud
  REPLICATED VALUES PLAINTEXT readiness highlight.
• Added near-cache RepairingTask, backup promotion, per-replica
  anti-entropy, hot-copy invalidation directory, and aggregate grid counters.
• Added release-gate tests for placement, rebalance, tombstones, replication,
  data protection, near-cache repair, failover, anti-entropy, hot-cache
  invalidation, transport wire safety, raft persistence seam, ADR presence, and
  metric cardinality.

Changes

• New module-level grid API exported from hydracache:
  ClusterReplicationStrategy, Replicas, EffectiveReplicationMap,
  ReplicationConfig, ReplicatedSlot, TombstoneTracker,
  PartitionReplicaVersions, HotCacheDirectory, and related helpers.
• New HydraCache methods:
  cluster_grid_counters, replication_config,
  replicated_value_security_posture, and record/set helpers for replication,
  under-replication, fence rejection, and tombstone repair-debt counters.
• New builder knobs:
  replicate_values, replication_factor, read_quorum, write_quorum,
  sync_backups, async_backups, max_replicated_entry_bytes,
  acknowledge_plaintext_replicated_values, and
  replicated_values_encrypted.
• HydraCache::client().connect() and HydraCache::member().start() validate
  the replication config before admission.
• hydracache-cluster-raft now has RaftLogStore and
  InMemoryRaftLogStore; the runtime Ready loop persists snapshot, entries,
  and hard state through that seam.
• hydracache-cluster-transport-axum has an explicit replication gate test for
  strict wire-version rejection and current-version encoded-byte roundtrip.
• hydracache-observability has cardinality tests ensuring grid metrics remain
  aggregate-only and high-cardinality detail stays in diagnostics snapshots.
• The workspace version is bumped to 0.41.0.

Verification

Focused gates:

cargo test -p hydracache --locked adr_presence
cargo test -p hydracache --locked topology_fence
cargo test -p hydracache --locked placement
cargo test -p hydracache --locked rebalance
cargo test -p hydracache --locked tombstone_replication
cargo test -p hydracache --locked replication
cargo test -p hydracache --locked replication_data_protection
cargo test -p hydracache --locked near_cache_repair
cargo test -p hydracache --locked failover
cargo test -p hydracache --locked anti_entropy
cargo test -p hydracache --locked hot_cache_invalidation
cargo test -p hydracache --locked fault_injector_selftest
cargo test -p hydracache-cluster-transport-axum --locked replication
cargo test -p hydracache-cluster-raft --locked persistent_log
cargo test -p hydracache-cluster-raft --locked --features sled-log-store persistent_log
cargo test -p hydracache-observability --locked cardinality

Full gates:

cargo fmt --all -- --check
cargo check --workspace --all-targets --locked
cargo test --workspace --all-targets --locked
cargo test --workspace --locked -- --ignored
cargo test --doc --workspace --locked
$env:RUSTDOCFLAGS = "-D warnings"; cargo doc --workspace --no-deps --locked

Scope Boundary

The release still does not provide full production distributed-cache-grid
readiness. In particular, it does not add a production multi-node Raft storage
engine, durable replicated values that survive restart, transparent invalidation
from arbitrary DB writes, split-brain auto-merge, distributed transactions, TLS
termination, certificate management, or KMS integration.