HC3 - High availability cluster?

[winromulus 19]

Hi,

 

Is there a way to create a HC3 highly available cluster? By this I mean to use two or more HC3 units so if the primary fails, the secondary can take it's place?

I've searched the forums but can't find anything relevant.

My project requires close to zero downtime so I'm wondering if anyone has any information about how to have a second unit on standby.

 

Thanks

[tinman 1,025]

Maybe you can describe why you need such configuration first, becaue:

 

- physical connected loads are always working, no matter if gateway is there or not

- modules connected together via association are working always, one might get small delay (sometime 1sec, sometimes less) during action->reaction

- no, there is officially no way to get hot-swap / cluster system with single system HC3, you might however make backup of primary HC3 and restore it to replacement HC3 if breaks (wiht Fibaro Support help, when you call them, you might get this done immediatelly, if you sign support contract even 24/7)

- with master/slave configuartion there is however "replace slave" option, it is still beta however so can't tell you how good/bad is that part working

[winromulus 19]

Hi @tinman

 

To me having a single controller looks like a clear case of single point of failure. I would need to have a gateway up at all time to ensure that scenes run as expected in case of a hardware failure.

While the option with a Fibaro support contract (didn't even know that existed, any links?) might be feasible if you're at the location at the time, if the device fails and you're not home this is a major issue. While I can "trust" my automation to manage my house, I cannot trust a single device to keep everything running, that's just a recipe for disaster. I agree about the modules linked by association but that's extremely limiting (for scenarios where the gateway is dead)

My hope was to have some sort of HA cluster where scenes/settings/connected devices are replicated and it would work either in an active-active mode (seems less likely) or at least active-passive mode. This way if the primary dies, the secondary can take over.

 

Just as a note, if this is not currently possible: Having a single point of failure for home automation especially when you have automated water/gas/heating/door locks seems like a major problem. While for non-critical stuff (shutters, coffee maker etc.) it might not be a problem, ending up in a scenario where you've left home and the whole thing crashes (picture leaving kids at home with a non-tech savvy person) just scares me. Of course I've taken measures to have manual overrides everywhere but most of us can agree that the overrides are not obvious or some of them require in-depth knowledge to operate.

[tinman 1,025]

 

 

2 hours ago, winromulus said:

While the option with a Fibaro support contract (didn't even know that existed, any links?)

 

i though you speak about building, well, for your house you will not able to pay that

 

2 hours ago, winromulus said:

While I can "trust" my automation to manage my house, I cannot trust a single device 

 

think different, single device can be as well:

- door lock -> not working -> you have to wait outside till help comes

- heating -> not working -> you will die in winter

 

2 hours ago, winromulus said:

My hope was to have some sort of HA cluster where scenes/settings/connected devices are replicated and it would work either in an active-active mode (seems less likely) or at least active-passive mode. This way if the primary dies, the secondary can take over.

 

actually, with backups and spare HC3, you can get the data restored within few minutes (up to next business day). 

 

 

2 hours ago, winromulus said:

a scenario where you've left home and the whole thing crashes (picture leaving kids at home with a non-tech savvy person) just scares me.

 

but oyu have no problem to leave them with electricity and heating turned on? Like every system, home automation might die.

Like with every other system someone have to fix when it breaks, simple. 

 

 

[winromulus 19]

Hi @tinman,

 

To be honest I don't agree with your assessment. It's one thing for a module to fail but completely other for everything (or, with the exception of some capabilities added by associated devices, mostly everything) to fail. Also usually, when it comes to professional devices, "waiting" with everything crashed until you have a fix is not acceptable to everyone. A simple analogy is the UPS: why buy an UPS when you can just wait for the power to come back? Why buy a backup generator? For some the downtime of a service is an acceptable matter, for me, personally, it is not and I'm pretty sure I'm not the only one (see UPS industry, for an analogy).

My expectation was for an automation platform that handles some pretty critical systems around the house to come in with the assumption that yes, as you put it, a device might fail, but you have measures in place to handle the situation without downtime, which gives you the time to, for example, call Fibaro Support or order new HC device or fiddle with backups. To put it plainly, I'd rather not "die in winter" but enjoy a cup of coffee while ordering a replacement for the failed device.

 

Anyway, I don't want to turn this topic into a debate about why redundancy is a good thing and why I prefer zero-downtime, but rather stick to the original question about how I can achieve high-availability, which it seems is not possible with the current version of HC3, unless someone else has additional input on the matter.

[tinman 1,025]

one of the reason why i prefer to create independant automation (heating runs on thermostates, which can control it manually when scenes are dead) is the "die in winter, 23:00 friday" situation

 

As i said, i haven't explored yet the "replace slave HC3x" option, this might be the only way to archive kind of redundancy.