Comments - Global Transaction ID
Content reproduced on this site is the property of its respective owners,
and this content is not reviewed in advance by MariaDB. The views, information and opinions
expressed by this content do not necessarily represent those of MariaDB or any other party.
In the section "Using Global Transaction IDs" you write "The replica remembers that master_use_gtid=slave_pos|master_pos". I suppose you mean current_pos and not master_pos ?
I've noticed that GTID Strict Mode causes issues in a master - master replication, when running an optimization task for all databases.
When executing the command "--auto-repair --optimize --all-databases" on server A. This results in the following error on server B: Last_SQL_Error: An attempt was made to binlog GTID 10-10-69958886 which would create an out-of-order sequence number with existing GTID 10-10-69958886, and gtid strict mode is enabled
Is this a bug or expected behaviour?
In Mariadb 10.1.9 - when I try to set SET GLOBAL gtid_slave_pos = 'X-XXX-XXX';
my server crashes. I have a galera cluster setup with mariadb version 10.1.9.
Similarly - i want to have asynchronous replication between two galera clusters - the moment I start slave on one of the nodes - it crashes. I think the similarity is that whenever we touch mysql tables - its crashing.
What could be wrong here?
Please report this as a bug.
what is the behaviour when gtid_ignore_duplicates=OFF?
Given a replication domain is just by gtid_domain_id (looking at the earlier definition) is the server-id part of gtid taken into account with gtid_ignore_duplicates? (i.e. are 1-3-42 and 1-5-42 the same?)
How is the stream order affected by out-of-order parallel replication where the slave has its own slaves with log-slave-updates=1?
what is the behaviour when gtid_ignore_duplicates=OFF?
Then it is the DBA's responsibility to ensure that the same event will not arrive at the same slave through two different multi-source replication connections. Eg. if S1->S2->S4 and S1->S3->S4, then each S1 event could be applied twice on S4 unless something like ignore_server_ids is used on S4. Same as normal replication.
Given a replication domain is just by gtid_domain_id (looking at the earlier definition) is the server-id part of gtid taken into account with gtid_ignore_duplicates? (i.e. are 1-3-42 and 1-5-42 the same?)
Correct, 1-3-42 and 1-5-42 would be considered the same for the purpose of gtid_ignore_duplicates. So gtid_strict_mode=ON should be employed to avoid this.
How is the stream order affected by out-of-order parallel replication where the slave has its own slaves with log-slave-updates=1?
Stream order for transactions in the same replication domain is not affected, they will always be in the same order in the slave binlog as in the master binlog with respect to each other.
Transactions in different replication domains can appear in a different order with respect to each other in the slave binlog, compared to the order on the master.
Effectively, each replication domain is logged as an independent stream. GTID mode tracks the replication position separately for each stream.