v0.1.54

Performance

• 8% performance improvement in the connection pooler by optimizing our use of Tokio futures @ygxio
• Additional performance improvements by using Tokio futures correctly @sgrif
• Remove several unnecessary memory allocations @ygxio

Bugs

• Don't validate replica identity during the schema sync & data sync phases of resharding if caller does not intend to use logical replication to sync real-time data @bhargavtheertham-cb
• Extended protocol pipeline messages were incorrectly ordered in cross-shard queries @meskill
• Schema sharding shouldn't trigger the incomplete omnisharded writes check @jkaczman
• DISTINCT state was not reset between cross-shard queries in the same transaction @levkk

Features

• Make OpenMetrics host configurable with openmetrics_host @murex971
• Block direct-to-shard queries when SET pgdog.sharding_key doesn't match a mapped value in list-based and range-based sharding configuration @jkaczman
• Setting read_only on a user in users.toml will direct that user's traffic to replicas, if configured @jkaczman
• Partial support for PREPARE and EXECUTE (including sharding); the prepared statement can be only executed with EXECUTE at the moment and cannot be executed using the extended protocol @levkk
• Wait for health check to pass before allowing traffic on newly added databases in pgdog.toml @sgrif
• Optional FIPS mode (compile time flag: cargo build -p pgdog --release --features fips)

Testing

• Basic Haskell acceptance tests @levkk
• Ensure we only generate a node id when NODE_ID is set @meskill

Code quality

• Remove library target and only build binary target, allowing us to more easily detect dead code @sgrif
• Remove a bunch of dead code @sgrif
• Cleanup multi-shard and row decoder code @levkk

Full Changelog: v0.1.53...v0.1.54