Details
Description
Similar to previous Jira issue MDEV-34180
HOST: mysql, mysql 5.7.44 binlog_format=ROW
HOST: m1, mariadb 10.6 GALERA NODE replicating from HOST mysql, Using_Gtid: No (log file and position)
HOST: m2 mariadb 10.6 GALERA NODE
HOST: m3 mariadb 10.6 GALERA NODE
Run this command on mysql:
CREATE TABLE `tt_mem` (
|
`mem_id` bigint(20) NOT NULL,
|
`mem_description` varchar(100) DEFAULT NULL,
|
PRIMARY KEY (`mem_id`)
|
) ENGINE=MEMORY DEFAULT CHARSET=utf8mb4;
|
Error on m1:
2024-05-22 16:11:07 1 [ERROR] WSREP: Vote 0 (success) on 78cebda7-1876-11ef-896b-8a58fca50d36:2565 is inconsistent with group. Leaving cluster.
|
Error on m2 and m3:
2024-05-22 16:11:06 2 [ERROR] Error in Log_event::read_log_event(): 'Found invalid event in binary log', data_len: 42, event_type: -94
|
2024-05-22 16:11:06 2 [ERROR] WSREP: applier could not read binlog event, seqno: 2565, len: 482
|
OTHER COMMANDS THAT PRODUCE SIMILAR ERRORS
CREATE TABLE `aa_tab` (
|
`col1` int(11) NOT NULL,
|
`col2` varchar(100) DEFAULT NULL,
|
PRIMARY KEY (`col1`)
|
) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4;
|
rename table aa_tab to dd_tab;
|
alter table aa_tab add column col3 varchar(10) after col2;
|
drop table aa_tab;
|
grant select on *.* to `edward`@`%`;
|
Attachments
Issue Links
- duplicates
-
MDEV-32366 WSREP closes because of high traffic load and other replication
-
- Closed
-
-
-
- Closed
-
-
-
- Closed
-
Activity
The issue can be reproduced for mysql 5.7.44, 8.0.39 and 9.0.1.
The patch is tested for mysql 5.7.44, 8.0.39 and 9.0.1
The "event_type: -94" in the error message on nodes m2 and m3 refers to a GTID_EVENT. This is the event that begins all event groups (InnoDB transactions or DDL/GRANT/etc.), and exists in the MariaDB binlogs but not in the MySQL binlogs. So the node m1 would not have seen an event type -94/GTID_EVENT (It would see a corresponding MySQL BEGIN or GTID event), and presumably m1 generated a GTID_EVENT instead to send to nodes m2/m3. So it is possible that this difference in starting event for the event group triggered the bug on m2/m3.
(Just an observation, I have not analysed in detail what goes wrong in the Galera code).
Linking issue that look like they could be the same underlying root cause
I reported this issue.
Note that it also affects 10.11 - which is the version we use.
/Eirik
A workaround is to replicate from mysql to a stand-alone mariadb with log_slave_updates=ON then from the stand-alone replicate to the Mariadb Galera cluster. This works without error.
Thanks, the fix has been merged with the main revision: https://github.com/MariaDB/server/commit/c0b11e75fff9ea513cd6cccfe47837678cedb473