Tiny Enterprise-Level Home Network

A

asciirobot

Hello.

I will attempt to not babble too much and get right to it. If anyone is interested in babble I'm sure it'll come out in additional questions and answers :)

I'm planning a FreeBSD-based home network. I already have some hardware from what was originally going to be a Windows Server failover cluster so lets say I have two nodes already. I have other hardware collecting for a file server so I'm going to concentrate on the two nodes for this post.

What I'd like to do is build a FreeBSD cluster for my home running various services (I can list my current thoughts for this so far if relevant) running as close to what a FreeBSD-in-the-Enterprise set up would be. I think my options are HAST/CARP (but this doesn't seem to be 'failover'), Corosync and Pacemaker, ...something else?

I've been out of the FreeBSD wold for a long time and I'm not caught up yet with the reading I've been doing so far. What are your thoughts and suggestions? What am I missing? What else should I be considering?

Please and thank you :)
 
Why two? A single one with a UPS is likely to be more than reliable enough for a home network. The only downtime is likely to be from administrative operations, such as reboots for kernel upgrades, or changing configurations. I probably have no more than half hour of downtime every two weeks, which is 2-1/2 nines. Good enough?

For failover, I read about HAST a few years ago. It seems to be well developed, although I worry about it failing in edge cases. It explicitly refuses to handle split brain automatically, yet it has no mechanism for preventing it in case of network partition. I don't see how it interacts with ZFS, but maybe I just didn't look hard enough. And you still need a mechanism like CARP above to get the services to be unified and failing over. I wonder how much work it would be to get it set up correctly. And how as a home user you'd perform the necessary testing to be able to trust it.
 
ralphbsz said:
Why two? A single one with a UPS is likely to be more than reliable enough for a home network. The only downtime is likely to be from administrative operations, such as reboots for kernel upgrades, or changing configurations. I probably have no more than half hour of downtime every two weeks, which is 2-1/2 nines. Good enough?

For failover, I read about HAST a few years ago. It seems to be well developed, although I worry about it failing in edge cases. It explicitly refuses to handle split brain automatically, yet it has no mechanism for preventing it in case of network partition. I don't see how it interacts with ZFS, but maybe I just didn't look hard enough. And you still need a mechanism like CARP above to get the services to be unified and failing over. I wonder how much work it would be to get it set up correctly. And how as a home user you'd perform the necessary testing to be able to trust it.
Click to expand...
The why is easy, because I would like to.

I’ll have a look at the split-brain thing, I’ve come across mention of this but haven’t delved down this path. And my understanding is that CARP and HAST go hand-in-hand to failover services.
 
You must log in or register to reply here.
Back
Top