Galera Cluster as slave of MySQL/MariaDB master-master cluster

[Sergio Charrua]

Hello,

I have a Master-Master 2 nodes cluster with MySql5.6 running for about 2.5 years, where both clusters replicate data to each other. Nodes are identified as Node1 and Node2, and both have different server_id.

Due to some dependencies, we need to use MariaDB 10.5, so I created a new 3 nodes cluster. Nodes are identified as Node3, Node4 and Node5 and the all have the same server_id.

After dumping data from Node2, to a .SQL file, I loaded the .SQL to Node3 and this is OK.

I followed this (pretty good) article https://developer.epages.com/blog/tech-stories/how-to-migrate-data-into-a-galera-cluster/  

and set Node3 as slave of Node2. 

While Node4 and Node5 are stopped, Node3 replicates all the transactions from Node2 without any issues.

But if I start Node4 (while Node5 is still offline), it starts replicating data from Node3 (while Node3 is still receiving data from Node2), and after a while the Galera cluster get losts: on Node4 I get the following logs:

Apr 26 10:43:38 SIPDB04 mariadbd: 2021-04-26 10:43:38 2 [ERROR] Error in Log_event::read_log_event(): 'Found invalid event in binary log', data_len: 42, event_type: -94

Apr 26 10:43:38 SIPDB04 mariadbd: 2021-04-26 10:43:38 2 [ERROR] WSREP: applier could not read binlog event, seqno: 1232279, len: 115

Apr 26 10:43:38 SIPDB04 mariadbd: 2021-04-26 10:43:38 0 [Note] WSREP: Member 1(node2) initiates vote on 9ce34c76-a38f-11eb-8225-02b9187a0281:1232279,984da6543308f296:

Apr 26 10:43:38 SIPDB04 mariadbd: 2021-04-26 10:43:38 0 [Note] WSREP: Member 0(node1) responds to vote on 9ce34c76-a38f-11eb-8225-02b9187a0281:1232279,0000000000000000: Success

Apr 26 10:43:38 SIPDB04 mariadbd: 2021-04-26 10:43:38 0 [Warning] WSREP: Received bogus VOTE message: 1232279.0, from node 826d13f0-a672-11eb-b56b-8b095d1d2aa6, expected > 1232293. Ignoring.

After that I can't query Node4.

And I can't shutdown Node4, as it gets ages to stop Mariadb (in fact, I had to "kill" the process).

And on Node3 the slave status is Ok, seems to be up without any error, but it doesn't replicate anymore. I have to "kill" MariaDB on Node3 (a systemctl stop doesn't work) and restart MariaDB and slave starts replicating from Node2 once again.

What am I doing something wrong here? Is MariaDB Galera not supposed to support replication from Mysql, while copying data to other Galera nodes?

Here is my MariaDB Galera configurations (/etc/mt.cnf.d/server.conf):

[mysqld]

log_slave_updates=1

[mariadb-10.5]

bind-address=0.0.0.0

binlog_format=ROW

default-storage-engine=innodb

innodb_autoinc_lock_mode=2

innodb_locks_unsafe_for_binlog=1

wsrep_on=ON

query_cache_size=0

query_cache_type=0

datadir=/var/lib/mysql

innodb_log_file_size=100M

innodb_file_per_table

innodb_flush_log_at_trx_commit=2

wsrep_provider=/usr/lib64/galera-4/libgalera_smm.so

wsrep_cluster_address="gcomm://10.19.139.10,10.19.139.11,10.19.139.12"

#wsrep_cluster_address="gcomm://"

wsrep_cluster_name='galera_cluster'

wsrep_node_address='10.19.139.10'

wsrep_node_name='node1'

wsrep_sst_method=rsync

wsrep_sst_auth=db_user:admin

server_id=3

this configuration is identical on Node4 and Node5 (only difference is the wsrep_node_address and wsrep_node_name parameters)

Any clue? 

Does any have a Galera Cluster slave of a MySql/MariaDB Master replicating data up and running in production? 

Thanks in advance!

Sergio