Who steered you to using an R-Pi these days? You're going to want something a bit better, especially in the RAM department. Like others mentioned, I would bet it's failing and restarting because R-Pis just aren't going to cut it for a node option anymore.

I say this as someone who was using an R-Pi Umbrel as my node for a few years and finally broke down and upgraded to a Start9 last year because of all the issues I was having. If I didn't restart the node every couple days, it would simply stop syncing new blocks. When I first got it, my Umbrel took over a week to sync the chain, and it was a good chunk smaller then (May 2021). When I got my Start9 last year, it synced the chain overnight. RAM makes SO MUCH difference, it's ridiculous.

Reply to this note

Please Login to reply.

Discussion

a R-pi with 8 or 16g ram won't be sufficient?

At the $120 pricepoint of a 16G R-Pi 5, there are better options. An 8G R-Pi 4 might be adequate, but I'd still opt for something better if you can swing it.

One of the main attractions of R-pi's is the energy efficiency. Costs what - 5x or10x less in kilowatt hours to run a node? Wondering what the trade offs of that are. Apparently having the node sync is one - but not sure how important that is. (I have yet to run a node)

Power usage is probably worth considering.

RAM is a huge factor when syncing the chain. It can be the difference between the process taking over a week vs taking a day or less, so long as your internet connection is not itself a bottleneck. That issue is only going to get worse for low RAM devices as time goes on.

Ideally, you should only have to sync the chain once, but there are situations that could require you to re-download and verify it.

There are two other reasons RAM is important, though. The first is the UTXO set, which must be stored both on your hard drive and in RAM so that the information is readily available for validating new blocks as they come in. This UTXO set takes up a few gigs of RAM all on its own, leaving less available for other processes.

Finally, 8G of RAM might be just fine if Bitcoin Core and maybe Electrs is all you are running on your node. However, if you are going to run a node, it shouldn't just sit on a shelf so you can point at it and say "That's my Bitcoin node." You should USE it! That means you will want to have other services running on it, too. Maybe a lightning implementation like LND or CLN, or a mempool explorer so you don't have to potentially dox what transactions you are interested in by using mempool.space's public website. Or maybe you also want to run a Bitaxe connected to DATUM on your node, so you are using your own node's mempool to construct block templates. You also have Alby Hub or LNBits for hosting lightning wallets and other lightning services. Not to mention running Jam for decentralized coin-joins, or RoboSats for purchasing non-KYC sats.

So, for all of these reasons, I would not be looking at "What's the bare minimum of RAM I can get away with?"

Maybe the better question is: what hardware is best for someone looking into using lightning and developing with it? but that's ONLY what the hardware will be for...and then also, what hardware is needed to manage a high volume of small transactions?

people are so busy answering the question of 'how to run a lightning node' there isn't a lot of answering of 'how to have a great lightning setup'

That's going to be beyond the scope of my knowledge, since I am not a dev. I would not have a very good idea of what would be the ideal setup for someone focused on that single use-case.

I don't think there are going to be any hardware constraints for handling a high volume of small lightning transactions, though. Transactions on lightning don't have much data footprint to speak of, unlike on-chain. This is because on-chain you have to keep a copy of EVERYONE's transaction data, not just your own, while lightning you are only concerned with the updates to your own channel states.

I have yet to find any good information on figuring out power consumption of various computer systems.

Best I seem to find is just people recording results like this

https://community.start9.com/t/start9-server-power-consumption/180

hard to make sense of that being comparable to a r-pi. would think considering they don't use ARM chips that it's more expensive...also utilization of more ram seems like it'd be more expensive. Trying to learn more, still haven't found good resources.