Hi, I have my NC running at home on an Intel NUC with attached QNAP since several years.
Becasue at work I have more and more to do with clusters and Loadbalancers, some questions came into my mind:
How can I setup NC to run in a Failover mode? That means, one server is running the active session, the other one as backup only. If the master should become offline, the backup can overtake. As Loadbalancer I am using Zevenet CE.
How can I then extend NC to run in a cluster? That means, that the webserver, loadbalancer and even data folder are stored on multiple systems, but work together. If one node fails, the other one will still be online. If both are active, incoming traffic is spreaded between both (round robin). Also here used Loadbalancer is Zevenet CE
Does anyone have any experience in that?
I think you are looking for Global Scale
I remember there was talk about high-availability and scalability of Nextcloud somewhere (search youtube). From my professional experience redundant systems are way more complex and at the end it’s often better to follow KISS (keep it stupid simple) is far better. If you have good backup/restore procedures it might be better to accept an outage of X hours if the system fails for some reason until you bring it back online rather spend X/2 hours every month updating/monitoring/managing redundant system only to avoid/reduce downtime…
If you are looking for education and do this just because it’s possible - yes otherwise I would recommend not to implement redundancy until you have to.
Thanks for the information - there is currently no real need, but it would be interesting to know how it works. I guess if you have 4000 customers a standalone server might be undersized …
The Nextcloud enterprise support. Especially if you look into integration in special environments and stuff like that, they have experience running such large setups. Some tried here to have distributed systems etc. but it was more to learn something…
If possible, you can split a setup. Sharing between users is still possible with federated sharing. Global scale is a bit cleaner, there you can probably use a common login and it distributes automatically to specific servers. There is not too much documentation about it.