Details
Description
When a SQL KILL statement is requesting a transaction to be aborted at the same time when the same transaction is chosen as a victim in the Galera transaction certification process, the server can hang.
There have been attempts to fix this problem earlier. A suggested fix for MDEV-18464 had been pushed and soon thereafter reverted because of issues. Another fix (which adds another field to THD, expanding the potential state space) was pushed to 10.4 and 10.5 in MDEV-21910, but it fails to prevent such hangs.
It seems possible that something related to this caused MDEV-17092, which I had worked around by changing the code in InnoDB.
Attachments
Issue Links
- blocks
-
MDEV-18874 Galera test MW-286 causes Mutex = TTASEventMutex<GenericPolicy>]: Assertion `!is_owned()' failed. assertion
- Closed
- causes
-
MDEV-24704 Galera test failure on galera.galera_nopk_unicode
- Closed
-
MDEV-24829 10.5.8 fails to startup on approx 10% of ubuntu focal deployments
- Closed
-
MDEV-24915 Galera conflict resolution is unnecessarily complex
- Closed
-
MDEV-25114 Crash: WSREP: invalid state ROLLED_BACK (FATAL)
- Closed
- is blocked by
-
MDEV-23536 Race condition between KILL and transaction commit
- Closed
- is caused by
-
MDEV-21910 KIlling thread on Galera could cause mutex deadlock
- Closed
- relates to
-
MDEV-18534 galera_sr.GCF-1018B: Test failure: mysqltest: mysqltest: Can't connect to local MySQL server; signal 6
- Closed
-
MDEV-23536 Race condition between KILL and transaction commit
- Closed
-
MDEV-24966 Galera multi-master regression
- Closed
-
MDEV-25518 mariadb crash with mysqld got signal 11
- Closed
-
MDEV-17092 ERROR: AddressSanitizer: use-after-poison around lock_trx_handle_wait_low
- Closed
-
MDEV-18464 Port kill_one_trx fixes from 10.4 to 10.1
- Closed
-
MDEV-23851 Galera assertion at lock0lock.cc line 655 because of BF-BF lock wait
- Closed
-
MDEV-24923 Port selected Galera conflict resolution changes from 10.6
- Closed
-
MDEV-25368 Galera cluster hangs on Freeing items
- Closed