The username/password that you enter in the replication agent dialog on your authoring server is a user account on your publish server. So you have to have set up that username/password set in the publish server's security set up.
So if you have set up a new user account on your author server for this you have to do the following steps:
- Change your replication agent configuration back to the default (admin/admin unless you changed the publish password during your install process).
- Activate the new system account you created on your author server - this will push that account out to your publish server.
- Change your replication agent configuration to use your new system account.
The one thing to remember is that if you spin up a new publish instance you have to make sure the system account gets deployed there.
I got the replication to work for pages, but the users don't appear to be syncing. On the Author machine, I deleted all of the fake users that are installed automatically and created my own groups. Isn't the Author machine supposed to replicate this as well or does it only replicate content? If the latter, do I really have to go to each publish machine to delete these users (I thought packages just added content - not removed)?
Did you delete the out of the box users before you had the replication agents working? Replication agents don't sync repositories, they are event based syndication agents, so if you deleted the users while the replicaiton agents weren't working they might not have been synconized.
The other possibility is that the event listern that does replication of delete doesn't listen for user objects. The delete event listener is selective - it doesn't automatically replicate all delete events (for example if you delete a compoent from a paragraph system that only gets replicated when the page is replicated. I am not sure about users to be honest.
As to your question about packages they will remove content based on filter definitions. So as an example if you create a package that has as a filter /content/site/en-us/section with no rules that is interpreted as include everything below that node. If you build that package on instance A and it only has one page in it - say /content/site/en-us/section/pageA. If you deploy the package to instance B and instance B has /content/site/en-us/section/pageB on it installing the package will add pageA and delete pageB.