|
Id User Host db Command Time State Info Progress
|
...
|
211332 user IP:33930 DB Killed 998 starting COMMIT 0.000
|
211371 user IP:41030 DB Killed 944 starting COMMIT 0.000
|
211389 user IP:35044 DB Killed 1162 starting COMMIT 0.000
|
211395 user IP:52656 DB Killed 1181 NULL set autocommit=1 0.000
|
211401 user IP:36886 DB Killed 1265 starting COMMIT 0.000
|
211414 user IP:41334 DB Query 1191 Commit INSERT INTO TABLE (
|
SSID,
|
ACTIVITY_DATETIME,
|
USER_RSID,
|
|
0.000
|
...
|
The above is a small snippet from SHOW PROCESSLIST. Pulling one of the KILLED transactions from SHOW ENGINE INNODB STAUTS which ran slightly earlier-
---TRANSACTION 4123705336, ACTIVE 904 sec
|
12 lock struct(s), heap size 1128, 6 row lock(s), undo log entries 5
|
MariaDB thread id 211332, OS thread handle 139648232093440, query id 202880013 IP user starting
|
COMMIT
|
Trx read view will not see trx with id >= 4123705336, sees < 4123702558
|
Customer retested workload on 10.5.18 Enterprise Server Galera Cluster and did not encounter this problem.
We worked with the customer to test a case where only a single Galera node was fed all traffic (write and read) and verified the behavior persists and only the node handling the read/write workload is affected.
|