Hi Phil,

On 8 March 2017 at 11:12, Phil Sweeney <launchpad@sweeneymail.com> wrote:

[...]
SET GLOBAL slave_compressed_protocol = 1;

On both master and slave. Then stop/start slave.

However, when I monitor the network usage, I see no reduction in network traffic. However, I tried compression via an ssh tunnel and saw a big reduction (ssh doing the compression, not MariaDB). So the traffic is definitely compressible.
[...]

I am not an expert in MariaDB Server and ssh compression but maybe the algorithm used by ssh is different than the one used by MariaDB Server, which would explain what you observe. So your replication workload would not "play" well with replication compression.

I would suggest setting up a test environment and try things that you would expect compress well (a big LOAD DATA with a lot of repeated data, or an multi-INSERT with a lot of repeated data). If you still see compression not working well, it might be a bug and you should open a JIRA on that (with all the details of your test).

Best regards,

Jean-François Gagné