In the binlog router I collect information about the number of binlog events received by the master and also sent to each connected slave. This is useful, especially if graphed as the numbers should be inline, except for a disconnection from the master or a slave.

I notice:

$ maxadmin show services > service
$ grep events service | grep -v minute | head -20
No. of binlog events received this session: 869934
Total no. of binlog events received: 1739868
...
No. events sent: 0 # one line per slave, this connection is a monitoring script so it has received no binlog events
No. events sent: 816245
No. events sent: 816245
No. events sent: 816245
No. events sent: 816245
No. events sent: 816245
No. events sent: 816245
No. events sent: 816245
No. events sent: 816245
No. events sent: 816245
No. events sent: 816245
No. events sent: 816245
$

The numbers don't match up exactly but that's because there's a continuous stream of events coming from the upstream master and also there's a short period of time while the master is up before the slaves connect to this server (after a maxscale shutdown).

However, looking at the numbers it looks clear to me that the number of binlog events received by the master is roughly double that sent to the slaves. Given there's no filtering (maxscale does not do this) it looks to me as if the reporting of the received binlog events is double counted on the master.

Please check and correct.