v0.1.17

New features

• Support search_path for schema-based sharding. Now you can SET search_path TO schema_a, public and all subsequent queries will be sent to shard matching schema_a.
• Support generating globally unique BIGINT identifiers using the Snowflake-like time-based algorithm. Usable with SHOW pgdog.unique_id currently. pgdog.unique_id() function will be added in the next release.
• Automatic role detection by fetching pg_is_in_recovery() from each database. Enable it by setting lsn_check_delay = 0 and role = "auto".
• Track idle in transaction clients and report the metric in OpenMetrics and SHOW POOLS (sv_idle_xact).

Bug fixes

• Client connections weren't updating their status to active correctly inside transactions.
• Schema detection wasn't always working for SELECT queries with multiple schemas.
• Schema sync was ignoring IDENTITY column constraint.
• Aggregate column count was incorrect. @mijoharas
• PREPARE/EXECUTE not working correctly when prepared_statements = "full".
• REAL/DOUBLE PRECISION columns were dropped in rewritten INSERT and UPDATE statements used in cross-shard writes. @phillipvanheerden
• NULL values weren't correctly routed to all shards.

New Contributors

• @phillipvanheerden made their first contribution in #630

Full Changelog: v0.1.16...v0.1.17

v0.1.16

What's Changed

• feat: --cutover flag to schema-sync for bumping sequence values during traffic cutover
• feat: setting for controlling connection cleanup
• feat: statement pooler mode
• feat: replica lag monitoring, automatic database role detection & failover
• fix: connection pool optimization
• fix: omnisharded tables routing wasn't detecting joins
• fix: idle in transaction state wasn't set correctly on clients @jaggederest
• feat: sticky routing for omnisharded tables
• fix: converge on direct shard routing when seeing multiple tables and using schema-based sharding

Full Changelog: v0.1.15...v0.1.16

v0.1.15

What's Changed

• Support for schema-based sharding.
• Fix: config reload was delayed by one transaction.
• Add connect_time and connect_count stats to SHOW STATS.
• Fix: incorrect average calculations in SHOW STATS.
• Dockerfile accepts Postgres version as build argument @scottjacobsen
• Support for partitioned tables in data-sync command for resharding.
• Force-enable query parser to get full features in non-sharded databases (parse_queries_enabled configuration option) @mijoharas
• Support update and delete statements in data-sync for resharding with logical replication.

New Contributors

• @scottjacobsen made their first contribution in #601
• @mijoharas made their first contribution in #605

Full Changelog: v0.1.14...v0.1.15

v0.1.14

What's Changed

• Split multi-tuple INSERT statements to send tuples to their respective shards @jaggederest
• Fix memory leak in network buffer and make it cancel-safe. Large queries and result rows were not properly deallocated.
• Add tls_client_required setting to force clients to connect with TLS @zxaos
• Bump tooling in container to Postgres 18 @mscrivo

New Contributors

• @mscrivo made their first contribution in #593
• @zxaos made their first contribution in #587

Full Changelog: v0.1.13...v0.1.14

v0.1.13

What's Changed

• Add server_error statistic to show how many errors we received from Postgres (e.g. syntax errors, missing columns, etc.)
• Add include_primary_if_replica_banned read/write split strategy.

Full Changelog: v0.1.12...v0.1.13

v0.1.12

What's Changed

• Support sharding key updates across shards @jaggederest
• Fix server cleanup when client disconnects mid transaction
• Schema sync fixes: abort if primary key is missing, don't sync the publication itself

Full Changelog: v0.1.11...v0.1.12

v0.1.11

What's Changed

• Prepared statements: don't cache anonymous statements, because they may contain parameters (looking at you, Rails...)
• Prepared statements: fix eviction logic leaving zombie statements with zero usage in the cache. Eviction logic went down from O(n) to O(k) where k is the number of statements exceeding cache limits
• Cheaper clone for Bind messages
• Add snakeoil certs to Docker container for easier dev/staging TLS

Full Changelog: v0.1.10...v0.1.11

v0.1.10

What's Changed

• EXPLAIN can now show PgDog routing decisions @jaggederest
• Schema sync syncs all tables, extensions and schemas
• TLS configuration is hot-reloadable now @jaggederest
• Support for cross-shard stddev and variance aggregate functions @jaggederest
• Integration test suite improvements @jaggederest
• Use SCRAM auth for client connections with TLS instead of falling back to MD5
• Fix rollback bug on unsharded, primary-only deployments
• Queries can be cancelled when shutting down PgDog, which prevents long, stuck queries across pooler restarts @KR-bluejay
• Support rolling back savepoints in sharded deployments
• Allow for 65k parameters in prepared statements @jaggederest

New Contributors

• @KR-bluejay made their first contribution in #524

Full Changelog: v0.1.9...v0.1.10

v0.1.9

What's Changed

• Handle DISCARD, DEALLOCATE and DISCARD ALL commands correctly in transaction mode @ryanahall
• Support for AVG aggregate function in cross-shard queries, with automatic query rewriting and "ghost columns" @jaggederest
• Code coverage tracking @jaggederest
• Postgres <-> PgDog SQL acceptance tests @jaggederest
• Database parameters & mirroring support for environment-based configuration @grk
• Allow to specify [admin] section in users.toml, and be encrypted-at-rest.

Bug fixes

• Fixed leak in connection pool caused by a race condition with RELOAD command
• Database banning now works solely off of health checks and only for read queries

New Contributors

• @grk made their first contribution in #525

Full Changelog: v0.1.8...v0.1.9

v0.1.8

What's Changed

• Manual primary/replica routing with /* pgdog_role */ comment.
• Forward empty queries to a shard using round robin, instead of intercepting. Works better when using extended protocol.
• Add server_lifetime setting to control how long server connections are allowed to exist before being re-created by the connection pool.
• Make configcheck command work correctly.
• Fix dry run mode ignoring query comments for manual routing.
• Fix SET queries comments for manual routing.
• Don't use host banning when only one target is configured in the load balancer. 🙏 @ryanahall
• Added more JDBC driver examples and tests.

New Contributors

• @ryanahall made their first contribution in #498

Full Changelog: v0.1.7...v0.1.8