MDEV-35635: START SLAVE UNTIL allows CHANGE MASTER TO options #3909

MohanadKh03 · 2025-03-21T14:11:54Z

*The Jira issue number for this PR is: MDEV-35635

Description

This commit separates master's options from slave options in sql_yacc.yy by replacing master_file_def with slave_until_file_def which only contain the options needed in the query:

START SLAVE UNTIL

These are the only options valid in this query
MASTER_LOG_FILE_SYM, MASTER_LOG_POS_SYM, RELAY_LOG_FILE_SYM and RELAY_LOG_POS_SYM and MASTER_GTID_POS

This is only a change in the symbols the START SLAVE command did not use any functionalty related to what was included in master_file_def

Release Notes

How can this PR be tested?

This can be tested using start_slave_until.test file which should test valid options queries and invalid ones
Invalid being anything that is not part of this
MASTER_LOG_FILE_SYM, MASTER_LOG_POS_SYM, RELAY_LOG_FILE_SYM and RELAY_LOG_POS_SYM and MASTER_GTID_POS

Basing the PR against the correct MariaDB version

This is a new feature or a refactoring, and the PR is based against the main branch.
This is a bug fix, and the PR is based against the earliest maintained branch in which the bug can be reproduced.

PR quality check

I checked the CODING_STANDARDS.md file and my PR conforms to this where appropriate.
For any trivial modifications to the PR, I am ok with the reviewer making the changes themselves.

bnestere

Hi @MohanadKh03 !

Thanks for submitting the patch with a test case already! I've left a few suggestions to the test, and then a couple more suggestions for the git commit message itself:

Please keep each line to 72 characters in length
Git commit messages don't support markdown, so you can remove/simplify the formatting a bit (though note the PR description does support markdown)
Can you re-phrase/extend the problem description to start off with a description that is more user-facing? That is, a typical user won't know what "master_file_def options" are, but you can describe the problem without using that phrase. Then later in the commit message, you can mention it (as it is good to still mention this for developers).

bnestere · 2025-03-24T18:08:35Z

mysql-test/main/start_slave_until.test

+# Clean up after tests.
+DROP TABLE t1;
+source include/rpl_end.inc;
+# ==== End of test ====


Can you echo the "End of test" line? And also end the file in a new line.

bnestere · 2025-03-24T18:10:42Z

mysql-test/main/start_slave_until.test

+
+connection slave;
+-- error ER_PARSE_ERROR
+start slave until master_log_file='$master_log_file', master_log_pos=$master_log_pos_1, MASTER_USE_GTID=NO, MASTER_DEMOTE_TO_SLAVE=1;


To use value substitution (using for variables set via the let mysqltest construct), you need to start the line with the mysqltest command eval. So this should be

Suggested change

start slave until master_log_file='$master_log_file', master_log_pos=$master_log_pos_1, MASTER_USE_GTID=NO, MASTER_DEMOTE_TO_SLAVE=1;

eval start slave until master_log_file='$master_log_file', master_log_pos=$master_log_pos_1, MASTER_USE_GTID=NO, MASTER_DEMOTE_TO_SLAVE=1;

Though to the complete test case, I don't think you actually need to use substitution at all, as you are just testing parsing options. So ultimately I'd say you could do something like

Suggested change

start slave until master_log_file='$master_log_file', master_log_pos=$master_log_pos_1, MASTER_USE_GTID=NO, MASTER_DEMOTE_TO_SLAVE=1;

--error ER_PARSE_ERROR

start slave until master_log_file='master-bin.000001', master_log_pos=4, MASTER_USE_GTID=NO, MASTER_DEMOTE_TO_SLAVE=1;

--error ER_BAD_SLAVE

start slave until master_log_file='master-bin.000001', master_log_pos=4;

and a few other options to test completeness.

bnestere · 2025-03-24T18:12:49Z

mysql-test/main/start_slave_until.test

+# MDEV-35635
+# Testing issues around invalid parameters and START SLAVE UNTIL behavior.
+
+# Include necessary configurations


You shouldn't need any pre-amble source for this test, as it is just testing parsing. I think really all that this test needs are the actual start slave lines, no set-up data is needed. Then, for a "passing" test, you can just set the expectation ER_BAD_SLAVE (as no slave is actually configured, but it proves the parsing has gone through)

bnestere

Hi @MohanadKh03 ,

Looks great, thanks! I very much appreciate the spelling out of the test description and git commit message :)

I'm going to pass review along to our replication QA engineer, @mariadb-SusilBehera , to give an extra set of eyes to your MTR test.

bnestere · 2025-04-01T21:29:08Z

mysql-test/main/start_slave_until.test

+#
+# Valid query options (using only slave options)
+#  
+


Here could you --echo to restate

ER_BAD_SLAVE: Dummy error which should be treated as a passing test as this is a parser test not a logical one

So when looking through the .result file, it is stated that the error is good?

bnestere · 2025-04-01T21:32:56Z

sql/sql_yacc.yy

+  | RELAY_LOG_POS_SYM '=' ulong_num
+    {
+      Lex->mi.relay_log_pos = $3;
+      Lex->mi.relay_log_pos= MY_MAX(BIN_LOG_HEADER_SIZE, Lex->mi.relay_log_pos);


You should be able to just use $3, like in MASTER_LOG_POS_SYM's definition. Also, can you reiterate the comment about what this MY_MAX operator is for, as is done with master_file_def

Suggested change

Lex->mi.relay_log_pos= MY_MAX(BIN_LOG_HEADER_SIZE, Lex->mi.relay_log_pos);

/* Adjust if < BIN_LOG_HEADER_SIZE (same comment as master_file_dif rule for MASTER_LOG_POS_SYM) */

Lex->mi.relay_log_pos= MY_MAX(BIN_LOG_HEADER_SIZE, $3);

bnestere · 2025-04-01T21:34:25Z

mysql-test/main/start_slave_until.test

+--error ER_BAD_SLAVE
+START slave UNTIL MASTER_GTID_POS = '0-1-1';
+
+# Clean up after tests.


There's no clean up to be done, so you can probably just remove this line altogether

bnestere · 2025-04-01T21:34:45Z

mysql-test/main/start_slave_until.test

+START slave UNTIL MASTER_GTID_POS = '0-1-1';
+
+# Clean up after tests.
+--echo ==== End of test ====


Suggested change

--echo ==== End of test ====

--echo ==== End of test main.start_slave_until ====

mariadb-SusilBehera

I've added just one more review comment.

mariadb-SusilBehera · 2025-04-02T08:55:20Z

sql/sql_yacc.yy

        ;

+slave_until_file_def:


MASTER_GTID_POS is missing here. Shouldn't we include an entry for this here?

Not really needed inside this one this is just for handling master/slave combinations for MASTER/SLAVE_LOG_FILE and MASTER/SLAVE_LOG_POS

but MASTER_GTID_POS_SYM is already in the slave_until rule

server/sql/sql_yacc.yy

Lines 8397 to 8400 in 2fe28c6

| UNTIL_SYM MASTER_GTID_POS_SYM '=' TEXT_STRING_sys

{

Lex->mi.gtid_pos_str = $4;

Lex->mi.is_until_before_gtids= false;

bnestere · 2025-04-22T18:53:04Z

To report on the status of this ticket, the work has been reviewed and accepted, and is waiting to be QA tested as part of the 12.1 sprint, which is planned to start on June 3.

Restrict START SLAVE UNTIL to slave options disallowing master options START SLAVE UNTIL was allowing both slave and master options, which could cause confusion and misconfigurations. This commit fixes that by restricting it to these valid slave options: - MASTER_LOG_FILE, MASTER_LOG_POS - RELAY_LOG_FILE, RELAY_LOG_POS - MASTER_GTID_POS The parser change replaces `master_file_def` with `slave_until_file_def` in `sql_yacc.yy`, making sure master options like MASTER_USE_GTID and MASTER_DEMOTE_TO_SLAVE aren’t allowed anymore.

svoj added the External Contribution All PRs from entities outside of MariaDB Foundation, Corporation, Codership agreements. label Mar 21, 2025

bnestere requested changes Mar 24, 2025

View reviewed changes

MohanadKh03 force-pushed the mdev-35635 branch from 173083e to c22ee85 Compare March 25, 2025 00:07

MohanadKh03 marked this pull request as ready for review March 25, 2025 00:47

MohanadKh03 force-pushed the mdev-35635 branch from 81901fb to 2fe28c6 Compare March 25, 2025 17:24

MohanadKh03 requested a review from bnestere March 26, 2025 00:48

bnestere approved these changes Apr 1, 2025

View reviewed changes

bnestere self-assigned this Apr 1, 2025

bnestere requested a review from mariadb-SusilBehera April 1, 2025 21:36

mariadb-SusilBehera reviewed Apr 2, 2025

View reviewed changes

MohanadKh03 force-pushed the mdev-35635 branch 2 times, most recently from 5b57a37 to 4e73f9e Compare April 2, 2025 21:50

bnestere added the Replication Patches involved in replication label May 28, 2025

bnestere force-pushed the mdev-35635 branch from 4e73f9e to 48f859a Compare May 28, 2025 19:10

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Uh oh!

MDEV-35635: START SLAVE UNTIL allows CHANGE MASTER TO options #3909

MDEV-35635: START SLAVE UNTIL allows CHANGE MASTER TO options #3909

Uh oh!

MohanadKh03 commented Mar 21, 2025 •

edited

Loading

Uh oh!

bnestere left a comment

Uh oh!

bnestere Mar 24, 2025

Uh oh!

bnestere Mar 24, 2025

Uh oh!

bnestere Mar 24, 2025

Uh oh!

bnestere left a comment

Uh oh!

bnestere Apr 1, 2025

Uh oh!

bnestere Apr 1, 2025

Uh oh!

bnestere Apr 1, 2025

Uh oh!

bnestere Apr 1, 2025

Uh oh!

mariadb-SusilBehera left a comment

Uh oh!

mariadb-SusilBehera Apr 2, 2025

Uh oh!

MohanadKh03 Apr 2, 2025

Uh oh!

bnestere commented Apr 22, 2025

Uh oh!

Uh oh!

	start slave until master_log_file='$master_log_file', master_log_pos=$master_log_pos_1, MASTER_USE_GTID=NO, MASTER_DEMOTE_TO_SLAVE=1;
	eval start slave until master_log_file='$master_log_file', master_log_pos=$master_log_pos_1, MASTER_USE_GTID=NO, MASTER_DEMOTE_TO_SLAVE=1;

	Lex->mi.relay_log_pos= MY_MAX(BIN_LOG_HEADER_SIZE, Lex->mi.relay_log_pos);
	/* Adjust if < BIN_LOG_HEADER_SIZE (same comment as master_file_dif rule for MASTER_LOG_POS_SYM) */
	Lex->mi.relay_log_pos= MY_MAX(BIN_LOG_HEADER_SIZE, $3);

	--echo ==== End of test ====
	--echo ==== End of test main.start_slave_until ====

	\| UNTIL_SYM MASTER_GTID_POS_SYM '=' TEXT_STRING_sys
	{
	Lex->mi.gtid_pos_str = $4;
	Lex->mi.is_until_before_gtids= false;

		;

		slave_until_file_def:

Uh oh!

MDEV-35635: START SLAVE UNTIL allows CHANGE MASTER TO options #3909

Are you sure you want to change the base?

MDEV-35635: START SLAVE UNTIL allows CHANGE MASTER TO options #3909

Uh oh!

Conversation

MohanadKh03 commented Mar 21, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Description

Release Notes

How can this PR be tested?

Basing the PR against the correct MariaDB version

PR quality check

Uh oh!

bnestere left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

bnestere left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

mariadb-SusilBehera left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

bnestere commented Apr 22, 2025

Uh oh!

Uh oh!

MohanadKh03 commented Mar 21, 2025 •

edited

Loading