[kumarpallav-pplx]

Problem

The warm-pool config reads in crates/libs/cluster/pool/src/store.rs normalized the release hash in SQL:

CASE WHEN r.hash LIKE 'sha256:%' THEN r.hash ELSE 'sha256:' || r.hash END AS content_hash

Template-manager now registers BLAKE3 template identities by default, so a blake3:<hex> release is rewritten to sha256:blake3:<hex>. That fails to parse as a ContentHash (digest for sha256 must have 64 lowercase hex chars, got 71). The rows are collected into one Result, so a single BLAKE3 release on any pool config fails the whole warm-pool reconcile pass. No template gets refill, promotion, or cleanup, and pools for BLAKE3 releases never get slots.

Fix

Select template_releases.hash as stored and let ContentHash::from_str decode it. It already accepts bare sha256 hex, sha256:-prefixed, and blake3:-prefixed values, so bare legacy rows keep working. The change applies to the four config reads that shared the rewrite:

• enabled_configs
• reconcilable_configs
• recently_disabled_configs
• delete_empty_disabled_autoscale_configs, reached through autoscale_configs, the first read of each pass

observed_demand keeps its bare/prefixed matching. Its first arm, r.hash = s.content_hash, already matches BLAKE3 hashes exactly.

Tests

• New test config_reads_decode_blake3_release_hashes seeds a pool config on a blake3: release and asserts that all four reads decode it unchanged.
  • Before the fix: Error: ContentHash(InvalidHexLength { algorithm: Sha256, expected: 64, actual: 71 }), the same error the controller logs.
  • After the fix: the test passes.
• cargo test -p pplx-space-cluster-pool --lib: 68 passed, including the existing bare and prefixed sha256 hash tests.
• cargo test -p pplx-space-cluster-controller-server --lib warm_pool: 69 passed.
• cargo fmt --all -- --check and cargo clippy -p pplx-space-cluster-pool -p pplx-space-cluster-controller-server --locked --all-features --all-targets -- -D warnings are clean.
• I have not run the full workspace suite locally; CI runs it.

Rollout note

Clusters that have already registered BLAKE3 releases on pool configs recover on the first reconcile pass after the cluster-controller picks this up. No data change is needed.