How many of you are running a rpi node?
Discussion
rpi "routing" node is a bad path, but for personal node it will be ok.
My primary concern is public ln routing, but even for regular core use - just because you *can* use a shit tier computer, doesn’t mean you *should*.
i did for almost 3 years, but now im not. so much happier this way too 😉
I have two rpi nodes running Umbrel. Constantly going down or offline but did manage to route sats sometimes. Spun up a Start9 node on my old thinkpad a few months ago, much more stable. Only problem is no Tailscale vpn solution.
Yes. RaspiBlitz, but really just running as a bitcoin node. I have electrs on there too, and mempool. and if it dies, I don't care. I can bring up all of that quickly enough elsewhere (on rpi or something else) to meet my needs. I don't custody lightning, but I also don't keep any material amounts of currency in lightning.
How frequently does it drop on you?
Not very often. maybe seen it drop twice in the last year. I have my mempool in a browser window on a monitor that I use for that, music player, sometimes bitcoin chart, but not much else, so I might notice if it was down. But I might not, if it recovers gracefully. I think it may have failed to recover once in the last year, and needed me to reboot it, or at least recovered after I rebooted it. So, stable enough for my purposes. It's there if I need to make a transaction with my Sparrow wallet. I can watch Tick Tock Next Block, one my favorite shows. Good deal.
Hm. You seem to be in the minority. Is there a UPS sitting behind it?
Nope. I think a big difference is that I am not worried about lightning services. lnd might be dropping out all over the place, but I wouldn't notice, unless it also drops electrs. I think lnd is technically supposed to be running on it, but there's no channels, zero balance in the ln wallet, no need for it to actually do anything.
So I had RTL installed on there. Opened it up, it says it can't connect to LND server, so chances are, LND is down. Sparrow is connecting to my private electrum server just fine, so that appears to be working, and balances are current. I even have an electrum wallet with some random small balances in it, and that shows up fine in Sparrow.
So, good enough for me. I don't need to run lnd on my node for anything I'm doing right now. If I did, maybe I would choose something better than an rpi.
I have quite a bit of experience with friends not being able to keep theirs powered without faults, so I like to harp on rpi’s because of that. Glad it’s not afflicting you. Wondering what the difference is though.
I’d be interested to know block validation time vs traditional CPUs too. I’m sure it’s still keeping up fine & maybe needs an extra 200% time to complete, but if that took over 45 seconds I’d be surprised.
Do you know how to look that up? Making a mental note to circle back to that later.
My biggest limiting factor when syncing was storage speed. I started with a USB drive that was really slow (hadn't thought that through at the time), and took almost a month to sync. Then decided to go to a M.2, and I payed attention to speeds advertised. I was able to do a full sync including my electrum sync in about 2 days.
After that, I guess as long as it can validate a block in less than 10 minutes, we're good!
Lol ya that’ll hurt. Flash drives and sd cards are gonna rek so many. Backup habits will be learned through pain with their failure rate too.
I was wondering which proposals will cause validation time to get nutty. So far from what I’ve seen it’s nbd. Just wondering tho.
my umbrel goes down every week/every other week
Have you figured out why? What’s your fix?
I have rpi4 with 8Gb of ram for context.
I think it's just not a powerful enough unit to handle the load of X number of apps on said device. from bitcoind, electrum server, wallet software, to lightning nodes. I just think it's weak hardware.
I usually notice when I check my mempool and it isn't synced. usually the umbrel is still on & running but some service (not sure which I haven't investigated that far), sort of halts things up. my guess is its bitcoind (running umbrel w tor), as I have noticed when my umbrel halts, I usually don't have any active peers for bitcoin to talk to.
to fix I ssh into my umbrel and reset it from the console, and usually everything js resynced within 20 minutes.
I don't use this to route lightning because it is too unreliable.
a longer term solution is using a slightly old thin client office computer with an i5 processor and 8-16Gb of ram. I think node software packages are good starter kits, but should probably be run on beefier hardware - or install just the essential tools for your node on a clean debian/ubuntu install
I currently have 4 nodes. 1 is a RPI4 4GB, the others are 8GB. The 4GB one is running MyNodeBTC and the others running Raspibolt. Most also running Nodeyez on top. Two of the nodes have UPS as does all the networking gear for stability.
This month I will be building a new node using refurbished Dell Optiplex for stability, performance and expandibility. Will migrate key economic functions to it. Some pi nodes will be converted to pruned nodes so I dont have to upgrade their drives this year or next (they all have 1TB SSD). Others will be decommissioned and probably repurposed into another project.
Great point about 1TB’s, it’s simply not enough space! Good for backups & raid arrays tho.
Are you going to install on bare metal? I suggest a hypervisor to sit in between, it helps with flexibility.
Rpi’s are not pleb friendly nor robust enough to be a suitable long term bitcoin node, they’re tinkering tools for DIY enthusiasts. I will die on this hill.
I will likely install without hypervisor. Can you elaborate on why you think that adds flexibility
When you have something like proxmox on metal instead of Ubuntu (even though it does have kvm ootb), you get to create/destroy purpose built isolated environments (vm or container). So say you fucked up your btcpayserver so bad that it is not functional - just delete the instance & start over, no worries about your bitcoind functionality since it’s separate.
Or, say you’re about to fuck around & find out with some complicated maintenance - you can take snapshots & backups that let you rollback quickly to minimize downtime and allow for recovery.
What if you want to drop the specs of your bitcoind instance after IBD to save on cpu/ram resources? No problem!
The optionality is powerful.
Theres certainly tradeoffs. I will consider it for isolation
