pg-sharding/spqr 3.0.0

on GitHub

We are excited to announce SPQR 3.0.0! All the goals outlined in the SPQR 3.0 Roadmap have been implemented — distributed and reference tables, distributed query support, and two-phase commit for DDL/migrations across shards.

There are no breaking changes between 2.9.x and 3.0.0. We are simply opening a new chapter for SPQR and moving forward.

This release covers all changes from 2.9.9 to 3.0.0, including intermediate patch releases: 2.9.10, 2.9.11, 2.9.12

What's Changed

JOIN Routing & CTE Support

• Fix JOIN ON clause routing and add tests by @reshke #2365
• Move tuple routing logic to separate function and add test for JOIN by @reshke #2408
• Support LEFT and RIGHT join with aux-values multishard by @reshke #2409
• Fix JOIN adjust logic & add tests by @reshke #2466
• Support JOIN CTE adjust for transitive aux values routing by @reshke #2467
• Support multiple AUX CTE with one common ancestor for JOIN adjust by @reshke #2468
• Fix CTE alias referencing by @reshke #2415
• Fix relation alias resolve and add regress test by @reshke #2543

Extended Protocol & Multi-Portal

• Partially support multi-portal binding by @reshke #2519
• Stub implementation for portal suspension by @reshke #2522
• Use ProcessOneMsgCarefully for Describe by @reshke #2442
• Remove custom processing in Execute by @reshke #2443
• Fix pgproto message copy for simple query & parse by @reshke #2492
• Greasify protocol negotiation in startup processing by @reshke #2608
• Respect result format code in virtual pg_is_in_recovery() by @reshke #2517

Two-Phase Commit (2PC)

• Simple happy-path error recovery in two phase commit by @reshke #2480
• Fix transaction deploy on ROLLBACK by @reshke #2481
• Support client feedback for 2pc commit flow by @reshke #2531
• Add __spqr__set_next_2pc_gid and use it in tests by @reshke #2536
• Fix two phase recovery utility function by @reshke #2574
• Fix 2pc tx state recovery by @reshke #2578
• Add isolation tests for 2pc failure recovery by @reshke #2482 #2485

Implicit Transactions

• Support implicit tx blocks by @reshke #2557
• Add notice message for router implicit tx feature by @reshke #2561
• Fix implicit transaction error recovery flow by @reshke #2564

Auto-Linearize for Multi-Shard DML

• Auto-linearize hazard multi-shard upserts by @reshke #2512
• Fix auto-linearize on INSERT INTO .. RETURNING by @reshke #2527
• Reimplement dispatch linearize as plan hinting rather than GUC injection by @reshke #2560
• Support correct command complete tag in multishard UPDATE by @reshke #2618 #2622
• Fix multishard insert command complete tag by @reshke #2626

Data Transfer & Balancer

• Allow context cancellation for data transfer workers by @reshke #2386
• Allow to cancel data transfer task in any execution phase by @reshke #2390
• Cancel data move task only on IMMEDIATE flag by @reshke #2456
• Recheck key range status for in-flight data chunk by @reshke #2569
• GUCify in-flight data access option by @reshke #2571
• Do not perform parallel transactions on both shards in copyData by @EinKrebs #2627
• Trace all data transfer goroutines in coordinator by @reshke #2385

Metadata & Key Range Management

• Metadata version management basics by @EinKrebs #2395
• Implement router rebootstrap through SQL by @EinKrebs #2398
• Update key range metadata in spqrguard by @EinKrebs #2424
• Maintain key range meta on rename by @EinKrebs #2594
• Separate spqrguard meta update from external writes lock in config by @EinKrebs #2611
• Add a check for an exact key match in qdb/etcd by @Pavloxes #2552

Console & SQL Enhancements

• HELP for register/unregister routers, misc commands by @Denchick #2367 #2371
• Enforce 80-char line limit in HELP by @Denchick #2453
• Deprecate ADD keyword, use CREATE instead by @Denchick #2038
• Implement LOCAL & GLOBAL show kinds by @reshke #2440
• Support multistatement SQL in console by @reshke #2541
• Add ALTER SYSTEM ROTATE for spqr-console log rotation by @reshke #2558
• Add SHOW SHARDS_EXTENDED command by @diPhantxm #2562
• Support ALTER SHARD in router with coordinator by @diPhantxm #2428 #2445
• Add discard count column to SHOW pools by @reshke #2615
• Add server_lifetime setting to backend rules by @reshke #2616
• Improve shard-not-found errors in console by @Minetoff #2175
• SQL part of GRANT & alter system re-bootstrap by @reshke #2381

Monitoring & Observability

• Prometheus metrics for router (MVP) by @xelavopelk #2497
• Reimplement RPS counting as part of statistics package by @reshke #2475
• Fix router tx/shard statement quantiles by @reshke #2493
• Add simple query log time on INFO level by @reshke #2532
• Add and use TRACE severity for some debug messages by @reshke #2533
• Write log asynchronously (router only) by @reshke #2491 #2501
• Downgrade unmatched TSA host log level from Error to Debug by @Denchick #2488

spqr-monitor

• Initial implementation for spqr-monitor by @EinKrebs #2566

Shard Management & Networking

• Handle hosts with AZ in shard hosts validation by @diPhantxm #2486
• Pass hosts' AZ to distributed mgr by @diPhantxm #2498
• Send host list with AZ on router sync by @diPhantxm #2579
• Add gRPC validation rules by @diPhantxm #2450
• Fix memqdb rebootstrap losing shards by @diPhantxm #2441
• Make shard replicas synchronous by @EinKrebs #2515
• Change synchronous_commit to remote_apply on shards by @EinKrebs #2521
• Pin client catalog queries to dedicated shard on setup by @reshke #2609

Error Reporting

• Modernize coding around error reporting by @reshke #2372
• Switch errComplexQuery to use spqrerror infrastructure by @reshke #2414
• Use SPQR-specific error format for general error report by @reshke #2433
• Report key range id as errdetail by @reshke #2423
• Set QUERY field in SPQR-specific error for simple proto by @reshke #2535

Configuration

• Support UUID HASH column type in distributions by @Denchick #2588
• Support disabling idle_in_transaction_session_timeout by @EinKrebs #2617
• Support no-op distribution column update by @reshke #2621
• Cache schema in relay and share between executions by @reshke #2524

Stability & Panic Fixes

• Fix panic in SHOW relation & refactor by @EinKrebs #2407
• Fix panic in multishard conversion by @reshke #2526
• Fix router startup panic by @xelavopelk #2494
• Fix panic: unexpected proto version message size by @diPhantxm #2537
• Fix panic & release goroutines on route obsoletion by @reshke #2577
• Fix/out of range binds panic by @Ledatu #2556
• Remove panic from stat collector by @reshke #2476
• Make ICP panic action to be os.Exit() by @reshke #2499

Code Quality & Linting

• Fix var-naming linting issues, parts 1–4 by @Denchick #2378 #2420 #2427 #2431
• Fix unchecked-type-assertion, parts 1–2 by @Denchick #2392 #2394
• Fix unexported-naming, receiver-naming, var-declaration, unnecessary-if lint issues by @Denchick #2377 #2412 #2432 #2402
• Replace golang.org/x/exp/slices with stdlib slices by @Denchick #2473
• Remove dead code by @reshke #2404 #2477 #2539
• Remove ShardMapping, use TopologyManager instead by @reshke #2437

Tests

• Extended protocol tests: Copy, Close, Execute(MaxRows), simple-in-extended, BEGIN+extended and more by @rkhapov #2379 #2387 #2388 #2397 #2406 #2416 #2417 #2438 #2542 #2582 #2583
• Tests: Oid Field Description by @saneczkab #2425
• Add unit test CmpRangesLessStringDeprecated by @Vinter-mach #2502
• Add unit test: WHERE MatchRow by @NeSawer #2540
• Mock ClientInfo, RuleRouter and test ListClients, ListBackendConnections, ListPools by @AnastasiaKostareva #2451
• Fix flaky feature tests by @xelavopelk #2439 #2444 #2458 #2464
• Fix feature test flakes by isolating auth fixtures by @Denchick #2573

Documentation

• Add manual resharding guide by @singl3focus #2435
• Fix LOCK KEY RANGE behavior description by @Denchick #2448
• Fix typo in get_started guide by @kolzuk #2586
• Add isolation tests README by @Denchick #2603

DevOps & CI

• Use content hash for shard image cache by @Denchick #2452
• Fix mismatched hashFiles inputs between feature_prepare and feature runs by @Denchick #2465
• Fix CI unhealthy container log collection by @Denchick #2597
• Fix regress test teardown in schema_qual.sql by @Denchick #2592
• Bump PostgreSQL version by @reshke #2546
• Bumps by @dependabot #2373 #2469 #2470 #2511 #2530 #2565 #2589 #2593 #2602 #2607

New Contributors

• @saneczkab made their first contribution in #2425
• @singl3focus made their first contribution in #2435
• @Vinter-mach made their first contribution in #2502
• @NeSawer made their first contribution in #2540
• @kolzuk made their first contribution in #2586
• @Pavloxes made their first contribution in #2552

Full Changelog: 2.9.9...3.0.0