Run gh-ost on master, replica, or neither?

[joshlawrence]

Hello!

I am new to gh-ost, but familiar with pt-online-schema-change. Normally, I would run pt-osc on a master server, so I assumed that gh-ost would be the same. But in looking at gh-ost's doc, it appears that I should run it not on the master or the replica, but on my client machine that is able to connect to both my master and replica. Is this correct?

Apologies for what I'm sure is a dumb question, but I wanted to confirm.

Thanks!

[susanzhang27]

👋 Running gh-ost on replica is the preferred method as to reduce the extra load on the master. You can run gh-ost on the primary as long as you don't write to it also. More information on the different use cases can be found here

[ghnp5]

Hi @susanzhang27
I don't have replicas - only a master.

If gh-ost uses the binlogs, why can't I be writing in the master?
I was considering using gh-ost for a few ALTER TABLEs I need to run against huge tables that are constantly being written, on a busy database.

I thought gh-ost was fully appropriate for this job? But looks like I shouldn't use it if I don't have replicas, then?

Thanks!