Ubuntu Rhel 7
Master Slave Cluster

Is there anything we need to adapt explicitly during installation?

RagulR: Thanks for the report!

What are the slave's @@global.gtid_slave_pos? I suggest to use CHANGE MASTER TO .. MASTER_USE_GTID=slave_pos unless you already do that.

In the earlier version, we had MASTER_USE_GTID=Current_Pos and the above mentioned issue didn't occur. But only with the latest version upgrade, we are facing this issue.

Any recent change has been done regarding this?

What will be the impact on changing to MASTER_USE_GTID=slave_pos ?

RagulR In this sort of failure, sorry, it's not enough to refer to any successful past without providing more details,
including those that I've just asked for.

Your slave error suggests that either "slave"@@global.gtid_slave_pos or "slave"@@global.gtid_binlog_pos has a gtid value that does not exist in the master binlog.
Could you please paste result of SELECT of the two as well as a pair of variables from the master:
"master"@@global.gtid_binlog_pos and "master"@@global.gtid_binlog_state ?

I think it's reasonable that we first explain what is happening to your slave currently, and after that answer the question of why it had worked before.

Thank you.

Thanks. Here are the details of the server's and master's pos & state. Is there any other details needed

Slave Details:

>  mysql --defaults-file=/etc/na/my-1.cnf -uroot -pmysql

Welcome to the MariaDB monitor.  Commands end with ; or \g.

Your MariaDB connection id is 22

Server version: 10.2.26-MariaDB-log MariaDB Server

Copyright (c) 2000, 2018, Oracle, MariaDB Corporation Ab and others.

Type 'help;' or '\h' for help. Type '\c' to clear the current input statement.

MariaDB [(none)]> select @@GLOBAL.gtid_binlog_pos;

+--------------------------+

| @@GLOBAL.gtid_binlog_pos |

+--------------------------+

| 0-1-1                    |

+--------------------------+

1 row in set (0.00 sec)

MariaDB [(none)]> select @@GLOBAL.gtid_binlog_state;

+----------------------------+

| @@GLOBAL.gtid_binlog_state |

+----------------------------+

| 0-1-1                      |

+----------------------------+

1 row in set (0.00 sec)

Master Detail:

> mysql --defaults-file=/etc/na/my-1.cnf -uroot -pmysql

Welcome to the MariaDB monitor.  Commands end with ; or \g.

Your MariaDB connection id is 75

Server version: 10.2.26-MariaDB-log MariaDB Server

Copyright (c) 2000, 2018, Oracle, MariaDB Corporation Ab and others.

Type 'help;' or '\h' for help. Type '\c' to clear the current input statement.

MariaDB [(none)]> select @@GLOBAL.gtid_binlog_state;

+----------------------------+

| @@GLOBAL.gtid_binlog_state |

+----------------------------+

| 0-2-976                    |

+----------------------------+

1 row in set (0.00 sec)

MariaDB [(none)]> select @@GLOBAL.gtid_binlog_pos;

+--------------------------+

| @@GLOBAL.gtid_binlog_pos |

+--------------------------+

| 0-2-980                  |

+--------------------------+

1 row in set (0.00 sec)

Now with the gtid states available it's clear that the master refused to feed the slave correctly.
As why it may have worked before, there are few options. E.g slave@@GLOBAL.gtid_current_pos was empty, unlike as it is currently.

Thanks, Elkin. But still, I don't understand why this change I could see only after MariaDB upgrade

Before the upgrade, I could see the value of gtid_binlog_pos as empty in slave but gtid_current_pos has some value.

But after upgrade the state is different. Is there any recent change has been done regarding this and that should adapt during installation because no explicit change has been made from our end except the MariaDB version change?

What will be the impact on changing to MASTER_USE_GTID=slave_pos?

Hi Elkin. Is there any information needed from our end.

> gtid_binlog_pos as empty in slave but gtid_current_pos has some value

This may be fine 'cos of the definition of gtid_current_pos. Please check with the docs that defines both.
I can't really comment on your snapshot, except that it really looks nonsensical.
Specifically to the slave, where gtid_current_pos < gtid_slave_pos which goes against the definition.

If you have a reproducible test case that creates such a state we would really appreciate for sharing it,
otherwise I don't see how we can be helpful, sorry.

To your question of 'recent changes', I can't point to any that may relate to your description.

To 'What will be the impact on changing to MASTER_USE_GTID=slave_pos', I suggest you to read the mariadb docs' GTID section. Normally, the gtid tracking by slave should base on 'slave_pos'. To a normal case I relate for instance one that the slave does not update its state on its own.

Hi Elkin, Thanks for the clarification. But still, we have some doubts regarding this.
As u said, I tried to reproduce the scenario as simple as possible by a shell script mariaTest.sh with the below my.cnf file.

[mysqld]

server-id=1

log-bin=binary-log.

binlog-ignore-db=mysql

binlog-ignore-db=performance_schema

from where I could see the difference in the output of gtid_positions in v10.2.18 where first 3 positions are null while in v10.2.27 it has 0-1-1.

As part of my analysis, I could see the above difference only after adding "log-bin=binary-log." properties in my cnf file. while without the property I could not see the above difference.
Confused about the behavior after adding this property.

Can you please clarify the above behavior? Why the position get affected after adding "log-bin" ?
Or Is there any mistake I am doing in the my.cnf file?

Hi Elkin, can you comment on the below behaviour
On my understand and to summarize the above issue,

Note that when the binlog is empty (such as on a fresh install or after RESET MASTER), there are no event groups written in any replication domain, so in this case the value of gtid_binlog_pos will be the *empty string*.

As per the above note found in https://mariadb.com/kb/en/library/gtid/#gtid_binlog_pos.
Binlog-pos should be *empty string on fresh install but why the value is set as 0-1-1 *