Would DA be compatible with multiple (NOT multisite) Drupal installations, provided these would all point to the same database ?
In other words, can each child site point to its own IP address ?

I am aware that part of the rationale for DA is to use a single Drupal installation to share among several sites. However, we are coming across a situation where the several sites may have have to reside on different servers: owners of chidlren sites may request that they host their own site.

There is also an issue of load-balancing: each child site may not in itself require a load-balancing solution, but the combination of all children sites on a single server may generate too much traffic.

Comments

agentrickard’s picture

I assume the answer is "yes" -- as long as they all point to the same database. But I have neither the knowledge nor the resources to test this assumption for you.

agentrickard’s picture

Status: Active » Closed (fixed)
1kenthomas’s picture

Note: You're only talking about "separate installations" in a technical sense-- really, the scenario would have to be "mirrored installations" on separate servers, with a database server or servers on the side; and the question would make little sense WRT: two drupal installs (with different codebases) on the same server.

In this case, load balancing aside, I'm not sure the above reflects practical concerns. The data mostly resides in the database; so to think that having a "Drupal install" on your own server, while the database resides elsewhere, is "self hosting" is at the very least deceptive.

For performance, you'd certainly want the webservers and the databases at the same physical location (or to share a dedicated VPN or such, which I assume is impractical...), else the cross-net delay is going to erase your load balancing concerns.

Looking back at your post, you're probably not thinking in the right steps: if a single server (webserver+filesystem+database) generates too much load for the sites, the next step would to be to move the db to a dedicated box. After that, probably optimization and memcache. After that, probably virtual servers in some kind of farm/fallover configurations. After that...

Practically, I don't know very many networks that get to the farm stage... so this is overthinking.