Using The Master For Delivery? Options

[Flo]

Hi everyone!

I'm currently setting up two servers with openx using distributed statistics. At the beginning, the load won't be so extremely high that we'd need many servers. So I'm wondering if we could use the master for delivery as well when the first delivery slave is becoming loaded. Would there be a point in that?

Currently I'm suspecting it won't really help because the master would log all impressions in the binlog and transfer them to the slave. The slave might ignore those impressions because it is configured to ignore the %data_% tables, but I guess the log would still be transferred, clogging the database connection.

If I'm wrong and we could use the master for delivery, please tell me, because it would save us one server.

Greetings,
Florian

[David Keen]

You are right in saying that the binlog will be transferred to the delivery server which will then ignore the data_tables. This doesn't happen when using distributed maintenance because the binlog is switched off before inserting the data into the master server.

Whether this data will cause load issues with just one server you would have to test to find out. I can't think of any specific problems this would cause though.

As always, make a backup before messing about with stuff and preferably test it in a staging environment first.