server_id is a global variable. In multi-source replication setups or for external replication,
server_id can be useful as a session variable. In that case a session replaying a binary log from another server would set it to that server’s id. That way binary log has the ultimate source server id attached to it no matter how many hosts it passes, and it would provide loop detection for multi-source replication.
This was implemented by introducing the new
pseudo_server_id variable. This variable, when set to non-zero value, will cause all binary log events in that session to have that
server_id value. A new variable was introduced instead of converting
server_id to have both global and session scope in order to preserve compatibility.
You should use this option at your own risk because it is very easy to break replication when using
pseudo_server_id. One special case is circular replication which definitely will be broken if you set
pseudo_server_id to a value not assigned to any participating server (ie., setup is
pseudo_server_id is set to
5). It is also possible to create a temporary table
foo, then change
pseudo_server_id and from now
foo will not be visible by this session until
pseudo_server_id gets restored.
Command Line: Yes Config File: No Scope: Session Dynamic: Yes Default Value: 0
When this variable is set to
0 (default), it will use the global
server_id value. Note: this is different from the setting the global
0 which disables replication. Setting this variable to non-zero value will cause binary log events in that session to have it as
server_id value. Setting this variable requires