Thanks for all your responses!
I will try to clarify the nature of the project and answer all the questions.
A friend of mine has bought an decomissoned lighthouse from the Swedish Maritime Administration that he is renovating. The lighthouse is located on an island off the coast and due to bad weather, ice and the dark northern winter is unaccessible for a few months a year.
The lighthouse has been dark since 2007, but a couple of years ago i constructed a simple solution with a solarpanel, a car battery, twighlight sensor and a 12v car led bulb to light it up during the dark hours. However, without correct permission it could only be a fixed light.
Now the idea is to seek permission to get the old character back and light it as an "occas" light, ie occasionally. This means that it could for example be dark due to technical problems.
Last year he got mains power back into the lighthouse and a new feeder cable was laid out to the island (due to a youth hostel and a bird observatory also on the island, open during summer) so power will not be an issue, unless a total blackout would occur.
The cheap car led-bulb will be changed to a 230v bulb with the right color temperature, but will not be as bright as the original bulb, seen in the background on the image above.
It's hard to make suggestions if we don't know the specs. I would expect a lighthouse to be dry and at least partly heated (by the light waste heat), so the conditions aren't that extraordinary. But as I haven't ever driven a lighthouse myself I may be completely wrong.
There is no additional heat in the lanternine except the small amount that the bulb would produce, that would not be enough to heat even the inside of the lens, unfortunately. So the surronding temperature could range from -30°c to +30°c...
The linear regulator on the UNO board "burns" the 7V (from 12V to the internal 5V) into heat, so the thermic stress is considerable. 7 - 7.5V are much better. Or use a board with a switching regulator (DC/DC converter) there the input voltage isn't that relevant.
Power supply: 12V or whatever for your light bulb; buck converter to supply 5V to the Arduino (why that 7-9V? You got that from outdated tutorials maybe? Next you tell us you may want to use the barrel jack for power!).
Ok, so your tip about the dc/dc converter sounds good. As I said, I'm new to this so tips like this is really valuable.
And yes, I was considering using the barrel jack!
Board: Nano or Pro Mini, properly soldered on a piece of perfboard. Much more reliable than the Uno as there are no wiggle connectors involved.
Thanks, sounds like a solid coice. I see there's even a version of the nano with temperature- and light sensors built in, wich would help since I need a new twighligt sensor to turn the light on at dusk.
That way you catch only the simplest hardware failures (which are quite seldom BTW). If you really think that a redundant solution is necessary (I have serious doubts) use a system where each board is able to detect the output status (for example, does the light have power) and react if it detects that the state doesn't reflect the expectations. Such systems quickly get very complex, often you need three systems to build a quorum. What requirements do you have? Is the rest of the system comparably redundant (what happens if the light bulb fails, what happens if you have no power, what happens if you have a short circuit, etc.)?
It is hard for me to say whether two boards increase reliability. The changeover has to be taken care of reliably, or that becomes the weak link. Having an extra board also means there's one more part that could fail - this is why modern aircraft have only two engines, less risk of an engine breakdown. It also implies that you expect the rest of the hardware to be more reliable than the Arduino board itself, which I think is unlikely. I'd guess it's the lens rotation mechanism, light bulb, and the battery in that order that are the most likely to fail. Maybe it's the lighthouse construction itself even that should be in that top 3, as that is what's exposed to the rigors of weather.
What is the worst case senario if all boards fail?
Unattended for months means the possibility of failure.
You all got the point there, the Arduino is probably not the weakest link in this project.
The power source may fail, the power supply to the board, the realy controlling the flashes or the lightbulb itself. I should probably worry about those things instead!
If a link in the chain would break, it's not a issue other than my pride and that the lighthouse might be dark for a couple of months, it has no real importance for the navigation at sea any more.