[MDEV-31133] MariaDB Slave crash and cannot restart Created: 2023-04-26  Updated: 2023-04-26

Status: Open
Project: MariaDB Server
Component/s: Replication
Affects Version/s: 10.4.28
Fix Version/s: None

Type: Bug Priority: Critical
Reporter: Maurice Gasco Assignee: Unassigned
Resolution: Unresolved Votes: 0
Labels: None
Environment:

Rocky Linux release 8.7 on 4 servers with MariaDB 10.4 master / slave replication
1 MariaDB Server 10.4.24 as master
1 MariaDB Server 10.4.24 as slave
1 MariaDB Server 10.4.27 as slave
1 MariaDB Server 10.4.28 as slave (the one that crash)

Attachments: Text File first-crash.log     Text File other-crash.log    
Issue Links:
Relates
relates to MDEV-23713 Replication stops with "Index for tab... Stalled

 Description   

Slave crash during replication and crash again and again at restart. I could restart with innodb_force_recovery=3, reset slave but mariadb crash as soon as I restart without innodb_force_recovery.
Downgrading to MariaDB 10.4.27 solved the problem.

attached is the log of the first crash and one restart crash

 Comments   
Comment by Marko Mäkelä [ 2023-04-26 ]

This looks like the MDEV-9663 family of bugs, caused by the InnoDB change buffer. Between 10.4.27 and 10.4.28, the only somewhat related change is MDEV-29905.

Recent vesions of MariaDB 10.6 or later should not crash on this type of corruption thanks to MDEV-13542. The change buffer was disabled by default in MDEV-27734.

Comment by Alice Sherepa [ 2023-04-26 ]

mgasco So on 10.4.27 the problem is not reproducible?
Could you please add SHOW CREATE TABLE where update was? Do you use generated columns/ indexes on them? Hash indexes?

Comment by Maurice Gasco [ 2023-04-26 ]

MariaDB [demotelereleve1]> show create table iotdevicegroup\G
Table: iotdevicegroup
Create Table: CREATE TABLE `iotdevicegroup` (
`SYSID` varchar(40) NOT NULL,
`SYSVERSION` int(11) NOT NULL,
`SYSCLASS` varchar(80) NOT NULL,
`SYSCRETIME` datetime DEFAULT NULL,
`SYSUPDTIME` datetime DEFAULT NULL,
`SYSREAD` varchar(40) DEFAULT NULL,
`SYSWRITE` varchar(40) DEFAULT NULL,
`IOTCONTEXT_ID` varchar(40) DEFAULT NULL,
`IOTPARENT_ID` varchar(40) DEFAULT NULL,
`IOTTITLE` varchar(80) DEFAULT NULL,
`IOTCODE` varchar(40) DEFAULT NULL,
`IOTFULLPATH` longtext DEFAULT NULL,
`IOTUSERACCESS_ID` varchar(40) DEFAULT NULL,
PRIMARY KEY (`SYSID`),
UNIQUE KEY `CIADC4B51BCAAB314` (`IOTCONTEXT_ID`,`IOTCODE`),
UNIQUE KEY `CIADC4B51B67D37D5B` (`IOTCONTEXT_ID`,`IOTFULLPATH`) USING HASH,
KEY `FKds1nfbfvtubuii929y2l93ojk` (`IOTPARENT_ID`),
KEY `CIADC4B51BA38212CB` (`SYSCLASS`),
KEY `CIADC4B51B8A76B809` (`SYSUPDTIME`),
KEY `FK8c4ub1ie75wvir8fquoc50oi0` (`IOTUSERACCESS_ID`),
CONSTRAINT `FK8c4ub1ie75wvir8fquoc50oi0` FOREIGN KEY (`IOTUSERACCESS_ID`) REFERENCES `sysperson` (`SYSID`) ON DELETE SET NULL,
CONSTRAINT `FKds1nfbfvtubuii929y2l93ojk` FOREIGN KEY (`IOTPARENT_ID`) REFERENCES `iotdevicegroup` (`SYSID`) ON DELETE SET NULL,
CONSTRAINT `FKkbkqjjfg1s15pld37wkxwijpt` FOREIGN KEY (`IOTCONTEXT_ID`) REFERENCES `iotcontext` (`SYSID`) ON DELETE SET NULL
) ENGINE=InnoDB DEFAULT CHARSET=latin1 COLLATE=latin1_swedish_ci
1 row in set (0.000 sec)

Generated at Thu Feb 08 10:21:32 UTC 2024 using Jira 8.20.16#820016-sha1:9d11dbea5f4be3d4cc21f03a88dd11d8c8687422.