MDEV-28278: mysqlbinlog GTID start-position respects server_id

[ChandanaRamakrishna]

Fix for MDEV-28278.

mysqlbinlog should respect server_id when evaluating GTID start-position.
Events from different server_id should be skipped until a matching
domain_id + server_id + seq_no is found.

This patch adds server_id validation to ensure behavior matches
replica GTID filtering.

Tested on MariaDB 10.11 build.

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.

---

¨Chandana seems not to be a GitHub user. You need a GitHub account to be able to sign the CLA. If you have already a GitHub account, please add the email address used for this commit to your account.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[ParadoxV5]

Thanks for the work!

Would you mind extending the scope of this PR to include MDEV-37231 – that is, mariadb-binlog should respect server_id even if the position is otherwise found.

Please also include a test or two to prove that the problem is corrected (and to warn future improvements if they’re breaking intended behavior).
MDEV-37231 has a concrete example that you can reference.
(The tests doesn’t have to be fresh new files – adjustments to the test for --start-position (if there is one) also meets the requirements.)

[ParadoxV5]

This approach is incorrect.

• It’s a manual loop next to and overriding a method call with a very similar purpose.
• It even works by bypassing the gtid_state_validator section.

The fix to the check should most likely be a part of (that is, inside) exclude() – why should --stop-position not respect server_id, only --start-position?

[ChandanaRamakrishna]

Thanks for the review! I will update the implementation to integrate the server_id check into the GTID filtering logic and extend the fix to cover MDEV-37231 as well. I will also add tests demonstrating the corrected behavior. I'll push an updated patch shortly.

…

On Mon, 9 Mar 2026 at 20:26, ParadoxV5 ***@***.***> wrote: ***@***.**** requested changes on this pull request. Thanks for the work! Would you mind extending the scope of this PR to include MDEV-37231 <https://jira.mariadb.org/browse/MDEV-37231> – that is, mariadb-binlog should respect server_id even if the position is otherwise found. Please also include a test or two to prove that the problem is corrected (and to warn future improvements if they’re breaking intended behavior). MDEV-37231 has a concrete example that you can reference. (The tests doesn’t have to be fresh new files – adjustments to the test for --start-position (if there is one) also meets the requirements.) ------------------------------ In client/mysqlbinlog.cc <#4760 (comment)>: > + /* MDEV-28278: ensure server_id also matches when evaluating GTID start position */ + if (!gtid_event_filter->exclude(&ev_gtid)) + { + + if (position_gtid_filter && position_gtid_filter->get_num_start_gtids() > 0) + { This approach is incorrect. It’s not only an extra check that overrides another process, but also bypasses the gtid_state_validator block if the override activates. The fix to the check should most likely be a part of (that is, inside) exclude() – why should --stop-position not respect server_id, only --start-position? — Reply to this email directly, view it on GitHub <#4760 (review)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/A3O75UBGOSYL5YWTX5RWZL34P3LTVAVCNFSM6AAAAACWLVWBN6VHI2DSMVQWIX3LMV43YUDVNRWFEZLROVSXG5CSMV3GSZLXHMZTSMJVHEYDAMZWGA> . You are receiving this because you authored the thread.Message ID: ***@***.***>