Details
-
Task
-
Status: Open (View Workflow)
-
Major
-
Resolution: Unresolved
-
None
-
None
-
None
Description
Hi everybody,
It would be nice to have log_slave_updates to be set per replication stream instead of globally.
Imagine these nodes in a setup and let's say the 'A' replication data is much larger then the 'B' replication data:
Server A1 -> Main master
Server A2 -> Main secondary
Sever B1 -> Secondary master
Server B2 -> Secondary slave
With MSR we can replicate everything from A1 for the 'main' data. So:
A1 -> A2, A1 -> B1, A1 -> B2.
We want to have A1 -> B2 so in case of a problem or high load on B1 the 'main' data on B1 still stays up to date.
We want to set up circular replication between B1 <-> B2 for high(er) availability. This creates a problem: We do not want log slave updates enabled on the data that comes in from A1 to B1 because that is a lot of arbitrary data. Also filtering it with replicate_ignore_domain or other filtering solutions seams like a complex solution to a simple problem. With filtering this redundant data would still be written to binlog and sent over the network to the slave.
If we could disable log_slave_updates for the A1 -> B1 stream and the A1 -> B2 stream we could set up circullar replication between B1 and B2 without problems.