In other words, can I have a single central database, that gets data from remote databases, but that does not know which remote databases exist until data from the remote database arrives?

This does not sound like "replication" to me. Is there another term I should use for this, or is it a type of replication?


No. replicate-do-db specifies the databases to be replicated.

However, I'd say the performance you're describing is better aligned with "clustering." You might take a look at MySQL cluster (the NDB engine), as it may meet part of your expectations.

  • Thanks. As you could tell by the wording of the question, I strongly suspected this was not feasible. Is there a specific name for what I want to do? – MJB May 10 '10 at 0:39
  • Take a look at my update. – Warner May 10 '10 at 1:03
  • I looked at your update and don't understand it. It appears that clustering is intended to decrease downtime and improve redundancy. In this case, to add a new site, it seems you need to modify config files and reboot the whole cluster. I don't think that will be acceptable, since I could have many remote databases that will need to sync, and would not want to have to reboot repeatedly. It seems that if I want to allow updates from unexpected sites (with proper authentication somehow), I will have to roll my own. Or did I miss something? – MJB May 10 '10 at 14:09

No. MySQL replication is one-to-many. A slave can only replicate from one master database, but a master can replicate to many slaves.

To do what you want, you will have to craft your own solution on top of MySQL.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.