Details
-
Bug
-
Status: Needs Feedback (View Workflow)
-
Major
-
Resolution: Unresolved
-
12.3.2
-
None
-
None
-
Ubuntu 24.04.4 LTS, kernel 6.8.0-100-generic, x86_64 (DigitalOcean droplet, NVMe). MariaDB 12.3.2-ubu2404-log. InnoDB with io_uring active, innodb_buffer_pool_size=48G, thread_handling=pool-of-threads.
Description
mariadbd crashed spontaneously with signal 11 in an InnoDB background thread (ib_tpool_worker) during sustained heavy write/DELETE load (bulk import pipeline: hundreds of concurrent connections doing INSERT/UPDATE/DELETE on a 620GB schema; the coredump shows other threads in Sql_cmd_delete::delete_from_single_table). No client query was involved in the crashing thread (Query: null, connection 0).
The crash tore a page in a 114M-row table. Subsequent crash recovery could not repair it and refused to start, leaving the server in a systemd restart loop until started with innodb_force_recovery=1:
2026-06-06 14:04:33 0 [ERROR] InnoDB: Unable to apply log to corrupted page 14734719 in file ./bigshopper_nl_be/product_entries.ibd
|
2026-06-06 14:04:33 0 [ERROR] InnoDB: Plugin initialization aborted with error Data structure corruption
|
2026-06-06 14:04:33 0 [ERROR] Plugin 'InnoDB' registration as a STORAGE ENGINE failed.
|
Crash stack (error log)
260606 14:03:21 [ERROR] /usr/sbin/mariadbd got signal 11 ;
|
Server version: 12.3.2-MariaDB-ubu2404-log source revision: 9f98f82b14a9b939834281672b6d0cf965db69a3
|
Thread pointer: 0x6221da25e8e0
|
/usr/sbin/mariadbd(my_print_stacktrace+0x30)[0x62219c98b6d0]
|
/usr/sbin/mariadbd(handle_fatal_signal+0x1d7)[0x62219c53c017]
|
/lib/x86_64-linux-gnu/libc.so.6(+0x45330)[0x77a408c45330]
|
/usr/sbin/mariadbd(+0xf7abbf)[0x62219c8c0bbf]
|
/usr/sbin/mariadbd(+0xf7c128)[0x62219c8c2128]
|
/usr/sbin/mariadbd(+0xf2d8b9)[0x62219c8738b9]
|
/usr/sbin/mariadbd(+0xf2e498)[0x62219c874498]
|
/usr/sbin/mariadbd(+0xf04553)[0x62219c84a553]
|
/usr/sbin/mariadbd(+0xf4ac35)[0x62219c890c35]
|
/usr/sbin/mariadbd(+0xf4ae81)[0x62219c890e81]
|
/usr/sbin/mariadbd(_ZN5tpool10task_group7executeEPNS_4taskE+0x6c)[0x62219c934eac]
|
/usr/sbin/mariadbd(_ZN5tpool19thread_pool_generic11worker_mainEPNS_11worker_dataE+0x7c)[0x62219c93507c]
|
Connection ID (thread ID): 0
|
Query (0x0): (null)
|
Kernel log: ib_tpool_worker[2166270]: segfault at 10 ip 000062219c8c0bbf sp 00006fa388b27340 error 4 in mariadbd (read of address 0x10 = NULL pointer + struct offset).
Not reproduced on demand; the same workload runs nightly and crashed once, 4 days after upgrading to 12.3.2 (on 2026-06-02). A 404MB systemd coredump of the crash is available and can be shared on request.