[MDEV-37662] Binlog Corruption When tmpdir is Full - Jira

XML

Word

Printable

Bug Category:
Can result in hang or crash
Release Note Summary:

Hide
For large transactions (i.e. those whose data exceeds the binlog_cache_size limit and spills into tmp files), provides a general performance improvement and fixes a bug that would lead to binlog corruption where the large transaction would have its GTID written into the binary log, without any additional transaction data, if the server's --tmp-dir was full.

Show
For large transactions (i.e. those whose data exceeds the binlog_cache_size limit and spills into tmp files), provides a general performance improvement and fixes a bug that would lead to binlog corruption where the large transaction would have its GTID written into the binary log, without any additional transaction data, if the server's --tmp-dir was full.
Epic Link:
Reduce Slave Lag
Sprint:
Q4/2025 Server Development

Customer is seeing the following error in the error log file

2025-09-01  3:03:18 10421228 [ERROR] mariadbd: Error writing file '/mariadb/bin/logs/bin_logs/bin_log' (errno: 28 "No space left on device")

But the disk usage shows the disk has ample space in the binary logs partition

2025-09-01-03:03:03,  /dev/mapper/vg_mdb-lv_mariadb_tmp,     27044,    25325,   527,     98%,  /mariadb/tmp,

2025-09-01-03:03:03,  /dev/mapper/vg_mdb-lv_mariadb_data,    1012773,  733478,  237799,  76%,  /mariadb/data,

2025-09-01-03:03:03,  /dev/mapper/vg_mdb-lv_mariadb_bin,     1012773,  338199,  633080,  35%,  /mariadb/bin,

Customer is running an ANALYZE TABLE job with sql_log_bin=0, which consumes more tmpdir space.

But for some reason, it is throwing errors related to the binary log directory.

Additionally, this has also resulted in binary log file corruption, causing replication errors on the slaves

2025-09-01  3:13:12 6 [ERROR] Unexpected break of being relay-logged GTID 1-1-9786113615 event group by the current GTID event 1-1-9786113616

2025-09-01  3:13:12 6 [ERROR] Slave I/O: Relay log write failure: could not queue event from master, Internal MariaDB error code: 1595

2025-09-01  3:29:26 6 [ERROR] Unexpected break of being relay-logged GTID 1-1-9786113615 event group by the current GTID event 1-1-9786113616

2025-09-01  3:29:26 6 [ERROR] Slave I/O: Relay log write failure: could not queue event from master, Internal MariaDB error code: 1595

duplicates

MDEV-27803 MariaDB binlog corruption when "No space left on device" and stuck session killed by client

Error rendering 'com.xiplink.jira.git.jira_git_plugin:git-issue-webpanel'. Please contact your Jira administrators.