proxysql

Commit Graph

Author	SHA1	Message	Date
Rene Cannao	b16d1f44ee	Fix review issues in FFTO E2E tests (#5516 ) Two critical issues identified during code review: 1. test_ffto_mysql_transactions-t.cpp: TAP plan count mismatch. kPlannedTests was declared as 28 but the actual assertion count is 34 (1 connect + 9 scenario1 + 12 scenario2 + 3 scenario3 + 9 scenario4). This would cause TAP to report "planned 28 but ran 34" and fail unconditionally. Fixed by correcting the count to 34 and cleaning up the misleading breakdown comment. 2. test_ffto_mysql_concurrent-t.cpp: undefined behavior on pthread_create failure. If pthread_create failed for thread i, threads[i] was never initialized, but pthread_join(threads[i]) was still called unconditionally. Fixed by tracking which threads were successfully created via a bool array and only joining those. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>	3 months ago
Rene Cannao	7949016f62	Register new FFTO E2E tests in TAP groups (#5516 ) Add the 6 new FFTO E2E TAP test binaries to groups.json so they are discovered and executed by the TAP test runner infrastructure. All new tests are assigned to the same groups as the existing FFTO tests: legacy-g4 and mysql84-g4. These groups provide the required infrastructure (running ProxySQL + MySQL backend with fast_forward user configuration). New entries: - test_ffto_mysql_bypass_recovery-t (bypass stickiness + isolation) - test_ffto_mysql_concurrent-t (10-thread concurrent sessions) - test_ffto_mysql_large_queries-t (large SQL text + rapid fire) - test_ffto_mysql_large_resultsets-t (10K+ rows + threshold tests) - test_ffto_mysql_mixed_protocol-t (COM_QUERY + COM_STMT_EXECUTE) - test_ffto_mysql_transactions-t (BEGIN/COMMIT/ROLLBACK + PS×20) Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>	3 months ago
Rene Cannao	e92d82b713	FFTO E2E TAP test: concurrent sessions (#5516 ) Add test_ffto_mysql_concurrent-t.cpp covering FFTO behavior under concurrent load with 10 threads. Part of the comprehensive E2E test suite for issue #5516. Test scenarios (13 TAP assertions): 1. Table setup with 100 seed rows for concurrent SELECT operations. 2. 10 concurrent pthreads, each opening its own MYSQL* connection with fast_forward=1. Each thread executes 50 SELECTs and 50 INSERTs using per-thread ID ranges (base_id = 1000 + thread_id * 50) to avoid primary key conflicts. Each thread reports ok/fail. 3. Aggregate stats verification after all threads join: - SUM(count_star) >= 1000 (10 threads × 100 queries each) - SUM(sum_rows_affected) >= 500 (10 threads × 50 INSERTs each) Thread safety considerations: - TAP's ok() is not thread-safe. All ok()/diag() calls from worker threads are serialized via a global std::mutex, following the pattern from test_ssl_fast_forward-3.cpp. - Each thread gets its own MYSQL* connection → its own MySQL_Session → its own m_ffto instance. No shared FFTO state between threads. - GloMyQPro->update_query_digest() uses internal locking, so concurrent stats updates from multiple FFTO instances are safe. - Connection retries (up to 3 attempts) handle transient load. - Aggregate polling waits up to 5 seconds for async stats. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>	3 months ago
Rene Cannao	438758ee51	FFTO E2E TAP test: bypass recovery and session isolation (#5516 ) Add test_ffto_mysql_bypass_recovery-t.cpp covering FFTO bypass stickiness, per-session isolation, and buffer boundary behavior. Part of the comprehensive E2E test suite for issue #5516. Test scenarios (9 TAP assertions): 1. Bypass stickiness — sends a small query (recorded), then a large query that exceeds ffto_max_buffer_size (triggers bypass), then another small query on the same connection. The post-bypass small query must NOT be recorded, confirming that ffto_bypassed is a permanent flag for the session lifetime. 2. Per-session isolation — while connection A is bypassed, a fresh connection B opens and executes a small query. B's query must be recorded, confirming bypass is per-session (each MySQL_Session has its own ffto_bypassed flag and m_ffto instance). 3. Recovery after close — closes the bypassed connection A, opens connection C. C's queries must be recorded normally, confirming new sessions start with ffto_bypassed=false. 4. Buffer boundary — with buffer=200 bytes, sends a query whose MySQL packet payload is exactly 200 bytes (pkt_len == buffer). Since MySQLFFTO::on_client_data() checks `pkt_len > buffer` (not >=), this should NOT trigger bypass and the digest should be recorded. Uses poll_digest_count() helper for async stat verification and dump_digests() for diagnostic output on failures. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>	3 months ago
Rene Cannao	cf30e02c30	FFTO E2E TAP test: mixed text and binary protocol (#5516 ) Add test_ffto_mysql_mixed_protocol-t.cpp covering FFTO behavior when alternating between text protocol (COM_QUERY) and binary protocol (COM_STMT_PREPARE/EXECUTE/CLOSE) within the same fast-forward session. Part of the comprehensive E2E test suite for issue #5516. Test scenarios (22 TAP assertions): 1. Text SELECT followed by prepared SELECT on same table — verifies both produce digest entries and aggregate correctly when the normalized SQL text matches. 2. Interleaved text and binary operations (text INSERT → prepared INSERT → text UPDATE → prepared SELECT) — validates clean FFTO state machine transitions between COM_QUERY and COM_STMT_EXECUTE command types. 3. Prepare → execute → close → re-prepare same SQL → execute — exercises the COM_STMT_CLOSE path which erases the stmt_id from FFTO's m_statements map, followed by a fresh COM_STMT_PREPARE that gets a new stmt_id. Verifies both executions share one digest with count_star >= 2. 4. Two prepared statements active simultaneously (INSERT + SELECT) with interleaved execution — validates that FFTO correctly maintains separate m_statements map entries per stmt_id and attributes queries to the correct digest. Includes helper functions exec_prepared_insert() and exec_prepared_select() to reduce binary protocol boilerplate. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>	3 months ago
Rene Cannao	ac926629f7	FFTO E2E TAP test: transaction patterns and prepared statements (#5516 ) Add test_ffto_mysql_transactions-t.cpp covering FFTO behavior within explicit transactions and repeated prepared statement execution. Part of the comprehensive E2E test suite for issue #5516. Test scenarios (28 TAP assertions): 1. BEGIN → 2× INSERT → COMMIT — verifies that each statement type within a transaction gets its own digest entry. The two INSERTs normalize to one digest with count_star=2, sum_rows_affected=2. BEGIN and COMMIT are tracked as separate digests. 2. BEGIN → INSERT → SELECT → ROLLBACK — verifies all 4 statement types are tracked independently. The SELECT within the transaction should see the uncommitted row (sum_rows_sent=1). ROLLBACK gets its own digest entry. 3. Prepared statement (COM_STMT_PREPARE + COM_STMT_EXECUTE) executed 20 times with different parameter values — verifies count_star=20 and sum_rows_affected=20. All executions share one digest because the SQL template is identical; FFTO tracks this via its internal m_statements map (stmt_id → original SQL text). 4. SAVEPOINT → INSERT → RELEASE SAVEPOINT — verifies that SAVEPOINT and RELEASE SAVEPOINT produce distinct digest entries alongside the intermediate DML. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>	3 months ago
Rene Cannao	c8e7ba4258	FFTO E2E TAP test: large queries and rapid-fire execution (#5516 ) Add test_ffto_mysql_large_queries-t.cpp covering FFTO behavior with large SQL query text and rapid-fire execution patterns. Part of the comprehensive E2E test suite for issue #5516. Test scenarios (13 TAP assertions): 1. Long INSERT with ~10 KB VALUES clause (400 tuples) within 1 MB buffer — verifies FFTO records the digest and tracks rows_affected correctly for large query text. 2. Long WHERE clause with IN list of 2000 elements (~12 KB query) — verifies query text extraction works at scale. 3. Query exceeding 1 KB buffer on fresh connection — triggers the client-side bypass check in MySQLFFTO::on_client_data() where pkt_len > ffto_max_buffer_size. Verifies no digest is recorded. 4. 100 identical SELECTs in tight loop — validates count_star accuracy under rapid execution; all 100 should aggregate into a single digest entry. 5. 10 distinct queries (same template, different literal values) in rapid succession — verifies they normalize to one digest with count_star >= 10. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>	3 months ago
Rene Cannao	d063ebe6de	FFTO E2E TAP test: large result sets (#5516 ) Add test_ffto_mysql_large_resultsets-t.cpp covering FFTO behavior with large result sets. This is part of the comprehensive E2E test suite requested in issue #5516. Test scenarios (15 TAP assertions): 1. SELECT returning 10,000 rows — verifies sum_rows_sent accuracy in the FFTO READING_ROWS state machine under sustained load. 2. SELECT with ~1 KB TEXT columns (50 rows) — validates row counting with large per-row payloads that stress buffer management. 3. Result set just under ffto_max_buffer_size (64 KB) — confirms FFTO stays active and records the digest when within the threshold. 4. Result set exceeding the 64 KB threshold — verifies server-side bypass triggers and no digest is recorded. 5. Bulk INSERT of 500 rows — validates sum_rows_affected tracking for multi-value INSERT statements. The test uses the established FFTO TAP pattern: admin connection for configuration, fast_forward=1 user, verify_digest() with polling loop, and verify_no_digest() for bypass scenarios. Buffer size is reconfigured mid-test with fresh connections to pick up new thresholds. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>	3 months ago
Rene Cannao	5d02ad5faf	Add basictests TAP group and sysbench to CI base image Part of the CI migration from jenkins-build-scripts to in-repo test/infra/ infrastructure (#5522). Adds: - test/tap/groups/basictests/ group (benchmark, chuser, failover tests) - sysbench to docker-base Dockerfile for benchmark tests Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>	3 months ago
Rene Cannao	7f5bbf0f19	Fix compilation error: add missing #include <cstdint> in MonitorHealthDecision.cpp The int64_t cast in can_unshun_server() requires <cstdint>, which was not included, causing build failure. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>	3 months ago
René Cannaò	8368c8d279	Merge pull request #5467 from sysown/copilot/feature-load-restapi-routes-config Load `restapi_routes` from config into runtime on startup	3 months ago
René Cannaò	4d2bcc303e	Merge remote-tracking branch 'origin/v3.0' into copilot/feature-load-restapi-routes-config # Conflicts: # test/tap/groups/groups.json	3 months ago
René Cannaò	cb8d0aa737	Merge pull request #5456 from sysown/copilot/add-ssl-tls-certificate-stats-table feat: Add stats_tls_certificates and stats_global tables for TLS certificate metadata and tracking	3 months ago
René Cannaò	da13ab59e5	Merge remote-tracking branch 'origin/v3.0' into copilot/add-ssl-tls-certificate-stats-table # Conflicts: # test/tap/groups/groups.json	3 months ago
René Cannaò	9781f5992d	Merge remote-tracking branch 'origin/v3.0' into copilot/feature-load-restapi-routes-config	3 months ago
René Cannaò	a616b34b95	Register test_tls_stats-t in groups.json (g4 groups)	3 months ago
René Cannaò	d1bb48e444	Register test_load_restapi_from_config_startup-t in groups.json (g4 groups)	3 months ago
René Cannaò	efedd93c92	Rename stats_global to stats_proxysql_global to avoid strstr substring collision stats_global is a substring of stats_mysql_global and stats_pgsql_global, causing the strstr-based dispatch to match incorrectly. Exclusion guards don't work either because a query referencing both tables would suppress the stats_proxysql_global refresh. Renaming to stats_proxysql_global follows the existing naming convention (stats_mysql_global, stats_pgsql_global) and eliminates the substring collision entirely.	3 months ago
René Cannaò	c3564af9cf	Fix timeout_ms validation and write path section name consistency 1. Validate timeout_ms against the CHECK constraint (100..100000000) and skip routes with invalid or missing values with an error message. Previously, omitting timeout_ms defaulted to 0, which silently failed the SQLite CHECK constraint and dropped the route. 2. Update Write_Restapi_to_configfile to emit 'restapi_routes:' instead of 'restapi:' so save-then-reload cycles use the preferred section name consistently.	3 months ago
René Cannaò	9850d9daeb	Fix stats_global substring collision and reload failure tracking 1. stats_global strstr match: add exclusion guards for stats_mysql_global and stats_pgsql_global to prevent spurious stats___global() calls and unnecessary ssl_mutex contention on every MySQL/PgSQL global stats query. 2. TLS reload failure tracking: set tls_last_load_ok=false and update tls_load_count/tls_last_load_timestamp on non-bootstrap reload failures. Previously, a failed PROXYSQL RELOAD TLS would leave stale SUCCESS status in stats_global.	3 months ago
René Cannaò	428324088b	Merge pull request #5520 from sysown/unit-tests-skip-proxysql Add SKIP_PROXYSQL for infrastructure-free test groups (unit tests)	3 months ago
René Cannaò	3aab0c280f	Fix SKIP_PROXYSQL path: run test binaries directly instead of proxysql-tester.py proxysql-tester.py requires a ProxySQL admin connection and Docker-specific env vars (TESTS_LOGS_PATH, TAP_ADMINPORT, TAP_WORKDIR, etc.) that don't exist when SKIP_PROXYSQL=1. Instead, discover test binaries from groups.json and execute them directly on the host with a simple pass/fail summary. Addresses review feedback from CodeRabbit, Copilot, and Gemini on PR #5520.	3 months ago
René Cannaò	ceff98eeb1	Add README.md for TAP test groups directory Documents groups.json format (one entry per line, compact arrays), group naming conventions, directory structure, and infrastructure-free groups.	3 months ago
René Cannaò	4c4736c73d	Add SKIP_PROXYSQL support for infrastructure-free test groups Unit tests run standalone against libproxysql.a with stubs and need no ProxySQL instance or backend databases. This adds a SKIP_PROXYSQL flag that test groups can set in their env.sh to bypass all Docker infrastructure setup and run tests directly on the host. Changes: - ensure-infras.bash: source group env.sh early; exit immediately when SKIP_PROXYSQL=1 - run-tests-isolated.bash: source group env.sh early; skip ProxySQL and backend verification; run proxysql-tester.py directly on host instead of inside a Docker container - New test/tap/groups/unit-tests/env.sh with SKIP_PROXYSQL=1 - groups.json: add 18 unit test binaries to unit-tests-g1 group - Documentation: SKIP_PROXYSQL in env var tables and new section for infrastructure-free groups in test/infra/README.md	3 months ago
René Cannaò	16529445bf	Merge pull request #5482 from sysown/v3.0-5473 Unit test infrastructure: Glo* stubs, init helpers, and smoke test	3 months ago
René Cannaò	abd1da861f	Merge pull request #5470 from sysown/private/multi-group-runner Multi-group TAP test runner and ClickHouse test fixes	3 months ago
Rene Cannao	60b5aff5c4	Fix tests not running after infrastructure setup in multi-group runner The heredoc in run-multi-group.bash was causing bash to exit prematurely after ensure-infras.bash completed, without executing the test commands. This was due to shell expansion issues with the heredoc structure. Changes: - Replace heredoc with temp file approach in run-multi-group.bash - Add pre-proxysql hook call to ensure-infras.bash for cluster setup - Create legacy/pre-proxysql.bash for cluster node startup - Update mysql84/pre-proxysql.bash with cluster startup calls - Fix coverage collection to work correctly in parallel execution	3 months ago
René Cannaò	660ed5a907	Merge pull request #5515 from sysown/v3.0-5499 FFTO: PgSQL CommandComplete tag parser extraction + unit tests	3 months ago
René Cannaò	ea16ebb8ad	Merge branch 'v3.0-5473' into v3.0-5499 Signed-off-by: René Cannaò <rene@proxysql.com>	3 months ago
René Cannaò	f1f2eaa91e	Merge pull request #5514 from sysown/v3.0-5498 Phase 3.10: PgSQL error classification + unit tests	3 months ago
René Cannaò	d872a392a6	Merge branch 'v3.0-5473' into v3.0-5498 Signed-off-by: René Cannaò <rene@proxysql.com>	3 months ago
René Cannaò	ec5a9f9cc5	Merge pull request #5513 from sysown/v3.0-5497 Phase 3.9: PgSQL monitor health decisions + unit tests	3 months ago
René Cannaò	8d7459e560	Merge branch 'v3.0-5473' into v3.0-5497 Signed-off-by: René Cannaò <rene@proxysql.com>	3 months ago
René Cannaò	d232834c1c	Merge pull request #5512 from sysown/v3.0-5495 Phase 3.7: MySQL error classification + unit tests	3 months ago
René Cannaò	9bccd2d1fa	Merge branch 'v3.0-5473' into v3.0-5495 Signed-off-by: René Cannaò <rene@proxysql.com>	3 months ago
René Cannaò	cffd85f0ff	Merge pull request #5511 from sysown/v3.0-5494 Phase 3.6: Backend variable sync decisions + unit tests	3 months ago
René Cannaò	f5ace60850	Address review on PgSQL error classifier (PR #5514 ) - Fix: 57014 (query_canceled) is now classified as REPORT_TO_CLIENT instead of FATAL (it's a normal cancellation, not a server crash) - Fix docstring: crash_shutdown is 57P02 (class 57), not class 58; class 58 is system/I/O errors - Add test for 57014 exception (+1 test, plan 25→26)	3 months ago
René Cannaò	468adb450e	Address review on MySQL error classifier (PR #5512 ) - Remove unused MYSQL_ERROR_CONTINUE enum value - Clarify retries_remaining parameter semantics in docstring	3 months ago
René Cannaò	81f77d9d5c	Merge pull request #5508 from sysown/v3.0-5492 Phase 3.4: Server selection algorithm extraction + unit tests	3 months ago
René Cannaò	b51f50b96b	Address review: fix NULL asymmetry in backend sync decisions (PR #5511 ) - BackendSyncDecision.cpp: treat asymmetric NULLs as mismatches (client=NULL + backend="bob" → SYNC_USER, not SYNC_NONE) - Tests: replace weak null assertions (a >= 0 always true) with specific checks for SYNC_USER/SYNC_SCHEMA on asymmetric NULLs, both-null → no sync, schema asymmetric null (+2 tests, plan 15→17)	3 months ago
René Cannaò	c40eed9405	Merge branch 'v3.0-5473' into v3.0-5492 Signed-off-by: René Cannaò <rene@proxysql.com>	3 months ago
René Cannaò	1274fb7798	Add comprehensive FFTO protocol unit tests — MySQL + PgSQL (#5499 ) 36 test cases covering both MySQL and PgSQL FFTO protocol parsing: MySQL length-encoded integers (11 tests): - 1-byte, 2-byte, 3-byte, 8-byte values - Truncated buffers, empty input MySQL packet building (9 tests): - Normal, large (1000 bytes), empty packets - Header validation (length, seq_id, payload integrity) MySQL OK packet parsing (2 tests): - affected_rows extraction (1-byte and 2-byte lenenc) PgSQL CommandComplete extended (7 tests): - INSERT with OID, large row count, zero rows - All 8 command types in one sweep, 9 DDL commands - Null-terminated payload (real wire format) Fragmentation simulation (6 tests): - Truncated lenenc (partial data) - Multi-packet stream building and header verification	3 months ago
René Cannaò	f07cd83d16	Extract MySQL protocol utilities for FFTO unit testing (#5499 ) New: include/MySQLProtocolUtils.h, lib/MySQLProtocolUtils.cpp Functions: - mysql_read_lenenc_int(): reads MySQL length-encoded integers from raw buffers (mirrors read_lenenc_int in MySQLFFTO.cpp) - mysql_build_packet(): constructs MySQL wire-format packets (3-byte length + seq_id + payload) for crafting test data	3 months ago
René Cannaò	a066763ee9	Add PgSQL error classifier unit tests (Phase 3.10, #5498 ) 25 test cases: connection errors (08xxx), transaction errors (40xxx), resource errors (53xxx), fatal errors (57xxx/58xxx), non-retryable (syntax/constraint/data), edge cases (null/empty), retry conditions (retries, in-transaction, non-retryable, fatal).	3 months ago
René Cannaò	e82bcb391b	Extract PgSQL error classification logic (Phase 3.10, #5498 ) New: include/PgSQLErrorClassifier.h, lib/PgSQLErrorClassifier.cpp Functions: - classify_pgsql_error(): classifies by SQLSTATE class — connection (08), transaction rollback (40), resources (53) = retryable; operator intervention (57), system error (58) = fatal; all others (syntax 42, constraints 23, data 22) = report to client - pgsql_can_retry_error(): checks retry conditions (retries left, not in transaction — PgSQL transactions are atomic)	3 months ago
René Cannaò	fe49380e3f	Add PgSQL monitor unit tests (Phase 3.9, #5497 ) 11 test cases: ping shunning (threshold, boundary, disabled), read-only offline (writer/reader HG combinations).	3 months ago
René Cannaò	78400743ff	Extract PgSQL monitor health decisions (Phase 3.9, #5497 ) New: include/PgSQLMonitorDecision.h, lib/PgSQLMonitorDecision.cpp Functions: - pgsql_should_shun_on_ping_failure(): threshold-based shunning (simpler than MySQL — always aggressive with kill_all) - pgsql_should_offline_for_readonly(): read-only server in writer hostgroup should go OFFLINE_SOFT Unshun recovery is already covered by MonitorHealthDecision.h can_unshun_server() (same logic for both protocols).	3 months ago
René Cannaò	bc4251d589	Add backend sync decision unit tests (Phase 3.6, #5494 ) 15 test cases: no sync needed, schema mismatch, user mismatch, user+schema (schema handled by user change), autocommit mismatch, multiple mismatches, null handling.	3 months ago
René Cannaò	ead2331fb9	Extract backend variable sync decisions (Phase 3.6, #5494 ) New: include/BackendSyncDecision.h, lib/BackendSyncDecision.cpp determine_backend_sync_actions() returns bitmask of needed sync: - SYNC_USER: username mismatch → CHANGE USER required - SYNC_SCHEMA: schema mismatch → USE <db> required (skipped if SYNC_USER set, since CHANGE USER handles schema) - SYNC_AUTOCOMMIT: autocommit state mismatch Covers both MySQL (7 verify functions) and PgSQL (2 verify functions) at the decision level — the core comparisons are the same.	3 months ago
René Cannaò	56eb121f12	Add MySQL error classifier unit tests (Phase 3.7, #5495 ) 19 test cases: retryable errors with/without conditions, non-retryable errors (access denied, syntax, table not found, gone away), offline retry with all blocking conditions tested individually.	3 months ago

... 2 3 4 5 6 ...

9275 Commits (mysqlx-plugin-impl) All Branches Search

9275 Commits (mysqlx-plugin-impl)

All Branches