proxysql

Commit Graph

Author	SHA1	Message	Date
copilot-swe-agent[bot]	14ac53379a	test: harden admin credential logging regression Co-authored-by: renecannao <3645227+renecannao@users.noreply.github.com>	3 months ago
copilot-swe-agent[bot]	b4cd4a198e	fix: redact admin credentials in set logging Co-authored-by: renecannao <3645227+renecannao@users.noreply.github.com>	3 months ago
copilot-swe-agent[bot]	b76b6b17a9	Initial plan	3 months ago
copilot-swe-agent[bot]	2fbdd7df59	Initial plan	3 months ago
Rene Cannao	e9b78ec572	Fix reg_test_3847_admin_lock for isolated CI runtime layout The legacy version of reg_test_3847_admin_lock assumed the old TAP execution model where everything effectively lived on 127.0.0.1 behind docker port mappings. In the new legacy-g2 infrastructure the test process and the primary ProxySQL run in separate containers, while the test still launches a second local ProxySQL instance inside the test container. That split exposed several test-side assumptions that were no longer valid and caused the child replica setup to fail before the actual deadlock coverage even started. The most important issue was how the secondary ProxySQL was started. The test passed a static proxysql_sec.cfg, but it did not give the child process an isolated datadir. On the new CI the child could therefore discover and reuse an existing proxysql.db from the default runtime location, which meant the tracked config file was not actually authoritative. The test would then wait for a replica on 127.0.0.1:26081 that never came up with the expected credentials or cluster topology. This change makes the secondary runtime explicit and self-contained. The test now creates a private runtime directory under test/tap/tests/reg_test_3847_node_datadir/runtime, generates the child config dynamically for the current environment, and launches the child with -D pointing at that runtime directory. The generated config keeps the local child replica on 127.0.0.1:26081/36081 while wiring the primary cluster peer to the real TAP admin endpoint discovered from CommandLine. That removes the stale single-host assumption and makes the test match how the isolated container environment is actually wired. The fix also cleans up endpoint usage inside the test body. Connections to the primary ProxySQL admin interface now use cl.admin_host/cl.admin_port instead of the generic frontend host, while the worker that talks to the locally spawned child explicitly targets 127.0.0.1:26081. This matters in the new CI because the primary lives in a different container but the child lives next to the test process. There was also a correctness bug in the thread bookkeeping. The previous code could overwrite a real worker error with success, and failed worker connections did not stop execution immediately. The worker status flags and launcher status are now atomic, connect failures return immediately, main-loop checks only treat positive errno values as failures, and shutdown waits report the live state instead of silently racing past it. Since this class of failure is expensive to diagnose remotely, the TAP diagnostics are intentionally much more verbose now. The test logs the primary and replica endpoints it is using, the generated config path, runtime directory, launcher capture log, internal ProxySQL log path, worker startup messages, periodic progress from the two background loops, wait-loop heartbeats, and shutdown progress. That should make future CI failures explain themselves from the TAP log instead of requiring ad hoc reproduction. Finally, add a local .gitignore entry for reg_test_3847_node_datadir/runtime/. That directory is generated by the test and contains transient databases, logs, and certificates for the spawned child ProxySQL instance; it should not be committed.	3 months ago
Rene Cannao	7384bea084	Fix reg_test_unexp_ping_pkt for isolated CI networking and Prometheus labels The reg_test_unexp_ping_pkt TAP failure on legacy-g2 was caused by multiple test-side issues that became visible after moving from host-local 127.0.0.1 execution with port mapping to isolated Docker containers on a private network. The first failure happened during the initial setup path: the test connected with ProxySQL root credentials and root host, but used cl.mysql_port instead of cl.root_port. In the old port-mapped layout this mistake could be masked more easily. In the isolated DNS-based setup used by legacy-g2, root/admin/frontend and backend ports are distinct and the mismatch causes an immediate connect failure. This change routes the setup connections through the correct ProxySQL root port and adds targeted connection diagnostics so future CI failures show the exact host, port, user, and MySQL error involved. The second failure was a latent prepared-statement bug in the test data loader. MYSQL_BIND.length pointed to a block-local variable that went out of scope before mysql_stmt_execute(), which led to intermittent or immediate client-side failures such as 'Client run out of memory' when inserting the large LONGTEXT payload used by the test. This change keeps the bound length alive for the full execution window and adds payload-size logging so the expensive setup phase is visible in TAP logs. The final failure was in metric validation, not in ProxySQL behavior. ProxySQL exports proxysql_mysql_unexpected_frontend_com_ping_total as a labeled Prometheus series, for example with {protocol="mysql"}. The test previously looked up the metric by exact key, which silently read zero from the parsed map and made the assertion fail even though ProxySQL was correctly detecting and handling the unexpected COM_PING packets. This change resolves the metric by prefix, records how many matching series were found, and logs the before/after values and delta explicitly. Verification: - rebuilt test/tap/tests/reg_test_unexp_ping_pkt-t - executed the test directly against the live legacy-g2 ProxySQL instance and observed ok 9 / exit 0 - executed the isolated CI-style path with TEST_PY_TAP_INCL=reg_test_unexp_ping_pkt-t via test/infra/control/run-tests-isolated.bash and observed PASS 1/294 : FAIL 0/294 This commit intentionally does not include unrelated orchestrator configuration edits that were already present in the worktree.	3 months ago
Rene Cannao	98c03d6fc6	Fix stale per-connection state tracking in set_testing-multi set_testing-multi batches 10 queries on the same MYSQL* connection, but it was still generating a fresh random connection index on every loop iteration. The test then used that random index for two different purposes that should have referred to the active batched connection instead:\n\n- logging conn_idx in the debug output\n- saving vars back into varsperconn\n\nThat meant 9 out of every 10 iterations could store one connection's accumulated expected session state into a different slot. When a later batch reused that slot, the test loaded stale or foreign expectations and reported a ProxySQL mismatch even though the backend and ProxySQL session state were correct.\n\nTrack the active connection index for the duration of each queries_per_connections batch and use it consistently when loading state, logging, and saving varsperconn. This makes the debug output trustworthy and prevents intermittent false failures such as the legacy-g2 set_testing-multi regression that appeared to implicate variable synchronization.	3 months ago
copilot-swe-agent[bot]	0357862590	fix: address code review feedback - use strdup for file paths, check ASN1_TIME_diff return, refactor dup cert loading, larger CN buffer Co-authored-by: renecannao <3645227+renecannao@users.noreply.github.com>	3 months ago
copilot-swe-agent[bot]	4f88740b20	feat: add stats_tls_certificates table and TLS tracking metrics - Add STATS_SQLITE_TABLE_TLS_CERTIFICATES definition with cert_type, file_path, subject_cn, issuer_cn, serial_number, not_before, not_after, days_until_expiry, sha256_fingerprint, and loaded_at columns - Add TLS tracking fields to GloVars.global: tls_load_count, tls_last_load_timestamp, tls_last_load_ok, tls_cert_file, tls_ca_file, tls_key_file - Update ProxySQL_create_or_load_TLS() to populate tracking fields on success - Register stats_tls_certificates table in Admin_Bootstrap - Add stats___tls_certificates() function that reads certs from disk at query time, extracts metadata (CN, serial, validity dates, SHA-256 fingerprint, days_until_expiry) and populates the table - Add TLS variables to stats___mysql_global(): TLS_Load_Count, TLS_Last_Load_Timestamp, TLS_Last_Load_Result, TLS_Server_Cert_File, TLS_CA_Cert_File, TLS_Key_File - Wire up query detection and refresh in GenericRefreshStatistics() Co-authored-by: renecannao <3645227+renecannao@users.noreply.github.com>	3 months ago
copilot-swe-agent[bot]	d1e51a2cf8	Initial plan	3 months ago
René Cannaò	07d43f627a	Merge pull request #5454 from sysown/v3.0-ci_84_mysql84 [WIP] CI changes	3 months ago
copilot-swe-agent[bot]	8e37bbaa52	Refine listener conflict validation Co-authored-by: renecannao <3645227+renecannao@users.noreply.github.com>	3 months ago
copilot-swe-agent[bot]	1dace933e2	Validate conflicting listener ports at startup Co-authored-by: renecannao <3645227+renecannao@users.noreply.github.com>	3 months ago
copilot-swe-agent[bot]	cccc926786	Initial plan	3 months ago
René Cannaò	8c4252814f	Merge branch 'v3.0-ci_84' into v3.0-ci_84_mysql84 Signed-off-by: René Cannaò <rene@proxysql.com>	3 months ago
Rahim Kanji	2765c3a9c7	Added mysql84-* group	3 months ago
Rahim Kanji	cac8dfaf2f	Fix charset_unsigned_int-t for MySQL 8.4	3 months ago
Rahim Kanji	305e6db0b2	Added destroy-infra.bash	3 months ago
René Cannaò	e61e30dbb6	Merge pull request #5448 from sysown/v3.0_fix_dup_entries_groups Removed duplicate entries in groups.json	3 months ago
Rene Cannao	b9ee38360c	Clean up set_testing debugging and improve connection tracking - Remove GTID-specific debugging from set_testing.h and set_testing-multi-t.cpp - Remove verbose [DEBUG] prefixes from test logging - Add thread ID and connection index tracking to correlate queries with connections - Log expected_vars, mysql_vars, and proxysql_vars for each query execution - Simplify check_session_track_gtids function (remove debug logging) This provides cleaner logging to diagnose which client connection triggers variable errors like the wsrep_sync_wait issue.	3 months ago
Rene Cannao	fbfe9025f7	Add SKILL.md with agent instructions for running tests	3 months ago
Rene Cannao	7b282a8c2b	Add ROOT_PASSWORD to infra-mysql57-binlog .env for test authentication Add ROOT_PASSWORD variable derived from INFRA_ID to support root user authentication in binlog tests. This enables tests like test_binlog_fast_forward-t and test_com_binlog_dump_enables_fast_forward-t to connect as root. The password is dynamically generated from INFRA_ID hash for consistency across MySQL and ProxySQL configurations.	3 months ago
Rene Cannao	8957dc99e4	Add debugging output to set_testing-multi-t test Added comprehensive debug logging to help diagnose session_track_gtids failures in set_testing-multi-t: - Log connection switching and reuse events - Track varsperconn state throughout test execution - Log all variable updates from test cases - Add detailed session_track_gtids comparison logging - Log ProxySQL internal session values - Add debug output to check_session_track_gtids function Also includes critical fix: save vars back to varsperconn after each test iteration to persist expected variable state across connection switches.	3 months ago
Rene Cannao	53a6f011d2	Fix infra-mysql57-binlog infrastructure for GTID binlog reader tests This commit fixes the infra-mysql57-binlog infrastructure to properly support GTID-based query routing using proxysql_mysqlbinlog reader containers. The infrastructure was migrated from the old CI to the new isolated CI structure. PROBLEM: The binlog tests (test_binlog_reader-t, etc.) were failing because: 1. ProxySQL could not connect to the proxysql_mysqlbinlog readers on port 6020 2. The mysql_servers table had gtid_port pointing to MySQL containers instead of the reader containers 3. Hostname resolution was broken - mysql1.${INFRA} resolved to MySQL containers for queries but also needed to resolve to reader containers for GTID tracking SOLUTION: 1. docker-compose.yml - Added Docker network aliases to reader services: - reader1 service now has alias 'mysql1.${INFRA}' - reader2 service now has alias 'mysql2.${INFRA}' - reader3 service now has alias 'mysql3.${INFRA}' This allows ProxySQL to connect to mysql1.${INFRA}:3306 for queries (MySQL) and mysql1.${INFRA}:6020 for GTID tracking (reader), using the same hostname but different ports resolved to different containers via Docker networking. 2. conf/proxysql/infra-config.sql - Fixed GTID configuration: - Set gtid_port=6020 for all mysql_servers (was incorrectly set before) - Added comments explaining the GTID reader architecture - Added 'root' user to mysql_users for mysqlbinlog tests that connect as root - Users: sbtest7, sbtest8, root (for binlog dump tests) 3. .env - Fixed default INFRA value: - Changed from INFRA=${INFRA:-.} to INFRA=${INFRA:-infra-mysql57-binlog} - This ensures proper hostname generation like 'mysql1.infra-mysql57-binlog' - The old value '.' created invalid hostnames like 'mysql1..' ARCHITECTURE: - MySQL containers (mysql1, mysql2, mysql3): Handle queries on port 3306 - Reader containers (reader1, reader2, reader3): Run proxysql_mysqlbinlog, connect to their respective MySQL instances, and expose GTID info on port 6020 - Network aliases allow same hostname (mysql1.${INFRA}) to resolve to different containers based on port: 3306→MySQL, 6020→reader TESTS AFFECTED: - test_binlog_reader-t (PASSED) - test_binlog_reader_uses_previous_hostgroup-t - test_com_binlog_dump_enables_fast_forward-t - test_binlog_fast_forward-t Related: legacy-binlog TAP group uses this infrastructure	3 months ago
Rene Cannao	784b666402	Enable proxysql_mysqlbinlog reader services and fix binlog infrastructure - Uncomment and enable reader1/reader2/reader3 services in docker-compose.yml - Add container_name and network aliases for reader services - Fix docker-mysql-post.bash to allow user replication (remove SET SQL_LOG_BIN=0) - Fix orchestrator configs (add orc2 and orc3) - Add new legacy-binlog TAP group for isolated binlog testing - Move binlog tests from legacy-g3 to legacy-binlog-g1 - Add missing sbtest users to infra-mysql57 proxysql config - Add extensive debugging to test_binlog_reader-t.cpp This enables GTID tracking via proxysql_mysqlbinlog readers on port 6020.	3 months ago
Rene Cannao	8f7c531c84	Update binlog infra for new CI and add to legacy test group - Update docker-proxy-post.bash to use radmin/radmin credentials - Add infra-mysql57-binlog to legacy test group infras.lst	3 months ago
Rene Cannao	6e5547f34c	Add infra-mysql57-binlog infrastructure for binlog reader tests This infrastructure is required for the test_binlog_reader-t test which verifies ProxySQL integration with proxysql_mysqlbinlog utility for GTID consistency. Key differences from regular infra-mysql57: - Uses proxysql/ci-infra:mysql57-binlogreader Docker image - Configures GTID port (6020) on mysql_servers - Uses hostgroups 1900/1901 (PREFIX=19) for writer/reader - Includes gtid_from_hostgroup in query rules for GTID tracking - Creates only sbtest7/sbtest8 users (specifically for binlog tests) - Includes orchestrator for replication management - MySQL servers configured with session_track_gtids=OWN_GTID Copied from: ~/jenkins-build-scripts/infra-mysql57-binlog	3 months ago
Rene Cannao	7c7def55e9	Add debugging to test_binlog_reader-t and add missing sbtest users - Add extensive debugging output to test_binlog_reader-t.cpp including: - Test description and purpose - Connection settings and credentials - Environment variable values - Detailed error diagnostics with troubleshooting steps - Additional logging in create_testing_tables() and perform_rnd_selects() - Add missing sbtest7 and sbtest8 users to infra-mysql57 proxysql config These users are required by test_binlog_reader-t and were present in the old CI infra-mysql57-binlog but missing in the new infra.	3 months ago
Rene Cannao	e582973ff8	Fix: Correct python33 shebang typo to python3	3 months ago
Rene Cannao	b91642c2c5	Fix: Update python shebang for regression test script Updated the shebang line in test/tap/tests/reg_test_3992_fast_forward_malformed_packet-pymysql-t.py from '#!/usr/bin/env python' to '#!/usr/bin/env python3'. This ensures the script correctly uses the 'python3' interpreter, resolving potential 'command not found' errors in environments where 'python' is not directly available or aliased to python2. The rest of the codebase appears to consistently use 'python3' shebangs.	3 months ago
Rene Cannao	06af3f6c61	Fix: Update python shebangs to python3 Updated shebang lines in several Python scripts from '#!/usr/bin/env python' to '#!/usr/bin/env python3'. This ensures that these scripts correctly use the python3 interpreter, which is standard in the environment and avoids 'command not found' errors for 'python'. Affected files: - docker/scenarios/repl1/test1.py - docker/scenarios/repl1/test1_.py - scripts/kill_idle_backend_conns.py - scripts/legacy/export_users.py - scripts/legacy/metrics.py - scripts/stats_scrapper.py - test/tap/tests/reg_test_3992_fast_forward_malformed_packet-pymysql-t.py	3 months ago
Rene Cannao	725575a913	Fix test_flush_logs-t for Dockerized and remote environments Reworked the log flushing TAP test to support environments where the test runner and ProxySQL are in separate containers: - Implemented 'Scheduler Hack': Since the test runner lacks docker CLI/daemon access, signaling SIGUSR1 is performed by injecting a temporary task into ProxySQL's own scheduler. This task uses /bin/sh to find the correct worker PID via 'pidof' and signals it internally. - Optimized Log Reading: Updated fn_get_rotations to prioritize the shared volume mount (/var/lib/proxysql) available in the runner, falling back to remote docker exec only if necessary. - Improved Path Resolution: Replaced the potentially infinite loop for locating ProxySQL root with a safe directory traversal. - Environment Support: Exported PROXY_CONTAINER in env-isolated.bash to allow tests to identify the target ProxySQL container. - Updated Orchestrator passwords in MySQL 5.7 infra to match current INFRA_ID derived credentials.	3 months ago
Rene Cannao	e96f02083e	Fix indentation in proxysql-tester.py	3 months ago
Rene Cannao	d2abd58d41	Add missing test tools to CI base image - Install mysql-shell (mysqlsh) required for regression test #3992. - Install mycli required for regression test #3247. - Install iproute2 (tc) required for regression test #3273.	3 months ago
Rene Cannao	00b87013cc	Migrate and localize proxysql-tester.py logic - Migrate test-scripts from jenkins-build-scripts to test/scripts/. - Refactor proxysql-tester.py to skip tests missing from groups.json instead of failing. - Ensure correct path resolution and dependency links in localized environment. - Disable automatic pre-hook execution in the Python tester.	3 months ago
Rene Cannao	3df9d77d20	RESTORE: Recover full README.md content and fix Step 0 placement	3 months ago
Rene Cannao	6fa2b3c868	Localize CI base image definition and update setup guide - Migrate proxysql-ci-base Dockerfile to test/infra/docker-base/. - Document base image build as a mandatory Step 0 pre-requirement. - Include --network host flag in build instructions for network stability.	3 months ago
Rene Cannao	35bed0813c	Update README.md with comprehensive examples and test filtering guide - Add detailed examples for single-test execution and parallel runs. - Document TEST_PY_TAP_INCL for regex-based test filtering. - Include a guide for creating custom test subgroups (e.g., mysql84-g5). - Clarify ProxySQL cluster management and safety policies.	3 months ago
Rene Cannao	4d85025544	Finalize robust and self-healing infrastructure initialization - Refactor all docker-compose-init.bash scripts with path discrimination logic. - Prevent directory creation for file-level volume mounts (.crt, .key). - Implement self-healing conflict resolution for transient SSL files. - Ensure project-wide container health checks before reporting readiness. - Verified successful cold start of multi-backend legacy group.	3 months ago
Rene Cannao	9a6e8dea3f	Restore test functionality and document cluster management - Restore missing documentation for TEST_PY_TAP_INCL and cluster variables. - Disable automatic pre-hook execution in proxysql-tester.py to ensure pre-configured env parity. - Fix hostgroup discovery in test_binlog_fast_forward-t.cpp. - Enhance diagnostics and introductory headers across multiple TAP tests.	3 months ago
Rahim Kanji	5772ae5ee8	Removed duplicate entries	3 months ago
Rene Cannao	a267287495	Hardened infrastructure initialization and Orchestrator stability - Dynamically patch INFRA and ROOT_PASSWORD in Orchestrator JSON configs. - Implement immediate failure and log dump if any container crashes during startup. - Add robust project-wide container health checks in docker-compose-init.bash. - Fix missing /var/log/mysql directory mount for MySQL 8.4 containers.	3 months ago
René Cannaò	3803f11077	Merge pull request #5381 from sysown/v3.0_pgsql_meta_command_describe_table Implement PostgreSQL \d tablename meta-command support (Admin interface)	3 months ago
Rahim Kanji	9ab9787c91	Addressed AI review feedback	3 months ago
René Cannaò	eb2bb39da5	Merge pull request #5446 from sysown/v3.0-build-improvements Build improvements for 3.1.6 and 4.0.6	3 months ago
René Cannaò	16443dc082	Merge pull request #5416 from sysown/v3.0_pgsql-copy-error-recovery-issue-5415 Handle late COPY protocol messages after session exits fast_forward mode - P0	3 months ago
Rahim Kanji	778d6d2978	Add multi-version testing across 14 PostgreSQL versions (16.x, 17.x, 18.x)	3 months ago
Rene Cannao	bd9e99d698	Fix GIT_VERSION assignment to preserve version string in recursive make calls	3 months ago
Rene Cannao	c5ccb0fb19	Fix GIT_VERSION logic to correctly respect exported tier variables	3 months ago
Rene Cannao	55d86feefb	Pass PROXYSQL31, PROXYSQLFFTO and PROXYSQLTSDB into the build container	3 months ago

... 6 7 8 9 10 ...

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

9275 Commits (mysqlx-plugin-impl)

All Branches