Navigating between cacti instances

[ericdives]

I was wondering if anyone else was using cacti in a multi-instance configuration, and how they might move between servers easily.

We are a large-ish university with a mostly homegrown MRTG driven stats collection system. Polling servers exist in every distribution node (3 campuses, 2 campuses with one server each, one campus with 7 or more, with around 2000 total devices being polled). Since it's not always clear which node a given building is out of, I'm trying to figure out some "easy" way for people using the graphs to search for and browse to a given host.

Our current system actually has a "tree" that exists on one host, but for a given piece of equipment uses a URL that directs the browser to the appropriate poller (which of course is running its own Apache instance).

If Cacti allowed one to add a graph tree item that was simply an external link, that would be doable - I could script the addition of that fairly easily (we have a central database as part of our homegrown system that knows about all of the equipment; each poller daily queries the DB to make the appropriate MRTG config files). But, alas, that's not how it works now, and even Superlinks doesn't offer that particular option.

Poking through the code, I'd be tempted to add the option for the tree creation routines to actually query other Cacti DBs and add their tree information in, but set the links up to link to the other host. Thing is, I wouldn't want to go through the effort if the community (or developers) doesn't have any interest in the option.

So - does anyone else have multiple cacti servers in their environment, and do you have something set up to move between them easily?

[X-dark]

Yes, I have done it with url rewriting on an apache frontend. You must sync some tables between your Cacti instances (graph trees, users, perms, ...) and have some discipline about what instance you are using for creating these replicated items (you want to do it on the master instance). But it works fine and is transparent for users.

You will need to put some SSO solution in place as well to let user authenticate once and for all instances.

With this solution I currently poll more than 1,000,000 data sources across 12 Cacti instances (well a little less instances as some are not currently polling anything).