4 ms·
Is netbox intended to be some kind of "source of truth"? I mean, if MyVM has today IP: 10.10.10.200, and tomorrow somebody change that IP, * should I expect N
by oriettaxx 3y ago
Is netbox intended to be some kind of "source of truth"?
I mean, if MyVM has today IP: 10.10.10.200, and tomorrow somebody change that IP,
* should I expect Netbox to change the IP assigned to MyVM?
Does Netbox do some kind of auto-discovery? or that's not its role?
Can Netbox ping my VM to know if they are UP, or that's not its role?
- jeffg-nbl 3y agoNetBox is indeed intended as a source of truth. It it tightly focused on being the very best DCIM + IPAM solution it can be, which means that discovery / reconciliation and assurance / monitoring are not a problem we're currently trying to solve. Other products do those things well, and can integrate with NetBox via our extensibility facilities including API, plugins, and scripts. In fact, we just announced partnerships with a couple of vendors that do those very things: https://netboxlabs.com/news/strategic-partnerships-reduce-adoption-barriers-for-network-automation/ https://netboxlabs.com/news/strategic-partnerships-reduce-ad...
- oriettaxx 3y agook, very good, thanks. So, (sorry I probably should read the docs, and I will, but I'm pretty curious now) the day we implement it, we have to plan to add all our data by hand, am I right? (IP, subnets, details on routes/routers, switches, login URI, notes, locations, ecc ecc..)
- midasuni 3y agoA source of truth should only be changed based on a deliberate controlled action. Your VM shouldn’t be changing IPs, if “somebody” changes it you have a chaotic network and you just want a network discovery tool. The change to the VM’s IP should be done through an auditable change process (like a pull request). If the VM doesn’t match the source of truth, the VM is wrong, not the source of truth. That PR process would also update your telegraf plugins to ensure that the new IP is being monitored etc. Net is won’t change the process, your automation (an ansible playbook perhaps) would do the change based on the information in the source of truth Now it’s possible you only have a couple dozen hosts and a handful of networks and thus your source of truth could be an inventory file, that’s a reasonable solution. When you have dozens or hundreds of switches and hosts in the thousands or more range, I would prefer to have a UI wrapped around that file though, with various links between different locations, switch ports, MDU outputs, etc, that’s where netbox comes in. Your grafana dashboard can take the physical location and tie in with your various monitoring (host and environment) to identify a problem in a specific part of your data centre quickly and reliably, as netbix knows what rack the host is in, what switch it’s connected to, etc