Support MariaDB VECTOR during CDC#4454
Conversation
🔄 Flaky Test DetectedAnalysis: The MongoDB container setup hit a "not primary" race condition where db.createUser ran before the replica set finished electing a primary after rs.initiate(), an environmental flake unrelated to the MySQL-focused branch. ✅ Automatically retrying the workflow |
Code reviewNo issues found. Checked for bugs and CLAUDE.md compliance. |
🔄 Flaky Test DetectedAnalysis: The flow/e2e package hit its 20-minute package timeout with no single hung test (longest-running was only ~2m) and no deadlock in the dump, with the still-running tests being unrelated PostgreSQL e2e cases rather than the PR's MariaDB VECTOR change — indicating the known-flaky, high-concurrency suite simply overran its wall-clock budget. ✅ Automatically retrying the workflow |
There was a problem hiding this comment.
LGTM, I'd change the PR title (which is preserved in the commit history) because the type is not only supported in MariaDB VECTOR but also in MySQL 9.0+, according to the tests conditions themselves, and they use the same format.
Correction on my comment: I see we treat a different underlying type for MySQL.
❌ 1 Tests Failed:
View the top 1 failed test(s) by shortest run time
To view more test analytics, go to the Test Analytics Dashboard |
🔄 Flaky Test DetectedAnalysis: The fully self-contained test TestRunPipeline_FilterStripsLines failed with "src failed: signal: killed" because its trivial local printf subprocess was SIGKILLed externally (OOM/resource exhaustion on the heavily-parallel runner), not due to any logic bug—and it failed in only 1 of 5 matrix jobs on an unrelated MySQL PR. ✅ Automatically retrying the workflow |
Maria binlog encodes vectors as MYSQL_TYPE_VARCHAR and gives us
stringwhile MySQL comes as[]byteResolves DBI-826