[MDEV-18587] Galera: Rolling upgrade: cluster data incosistency with SR on upgraded 10.4 nodes during drop database Created: 2019-02-15  Updated: 2019-02-19  Resolved: 2019-02-19

Status: Closed
Project: MariaDB Server
Component/s: Galera
Affects Version/s: 10.4.3, 10.3.13
Fix Version/s: 10.4.3

Type: Bug Priority: Major
Reporter: Shahriyar Rzayev (Inactive) Assignee: Teemu Ollakka
Resolution: Fixed Votes: 0
Labels: galera
Environment:

Ubuntu 18.04



 Description   

Upgrade scenario:

  • Start 5 nodes of 10.3
  • Upgrade sequentally from 10.3 to 10.4 starting from node5
  • For each node run random values of:

size = [10, 20, 30]
unit = ['ROWS', 'STATEMENTS', 'bytes']
# Run set global wsrep_trx_fragment_size=10;
query = "set global wsrep_trx_fragment_size={}".format(choice(size))
start_obj.mysql_sql_runner(basedir_new, socket.format(cnt), query)
# Run set global wsrep_trx_fragment_unit='ROWS'
query = "set global wsrep_trx_fragment_unit='{}'".format(choice(unit))
start_obj.mysql_sql_runner(basedir_new, socket.format(cnt), query)

But it happens even only with 'bytes' and size of 10.

  • on node3 for eg, create database and then try to drop:

MariaDB [(none)]> drop database sbtest;
ERROR 1235 (42000): Streaming replication not supported with binlog_format=STATEMENT
 
MariaDB [(none)]> show databases;
+--------------------+
| Database           |
+--------------------+
| information_schema |
| mysql              |
| performance_schema |
| sbtest             |
| test               |
+--------------------+
5 rows in set (0.001 sec)

As you see it is not dropped. But instead it is dropped from other nodes:

MariaDB [(none)]> show databases;
+--------------------+
| Database           |
+--------------------+
| information_schema |
| mysql              |
| performance_schema |
| test               |
+--------------------+
4 rows in set (0.001 sec)

Which leads to data inconsistency.


Generated at Thu Feb 08 08:45:14 UTC 2024 using Jira 8.20.16#820016-sha1:9d11dbea5f4be3d4cc21f03a88dd11d8c8687422.