The Lightning Network FAQ

[Wind_FURY]

Hey, just want to plonk a link to this topic here for anyone interested in running a LN node in the USA:
Devastating "Infrastructure" bill in US - contact your representatives

Apparently they want to treat node operators as brokers due to 'facilitating transactions'.
Maybe worth following if you're located there..

Kohas787 will be so happy that he can start spam-calling the nearest FBI office, and report that his neighbor is running a Lightning node from the basement, earning $5.00 in routing fees per week.

If only somene wrote a guide in how to run a Lightning node over the TOR network, https://stopanddecrypt.medium.com/running-bitcoin-lightning-nodes-over-the-tor-network-2021-edition-489180297d5

[1713922810]

1713922810

[1713922810]

1713922810

[1713922810]

1713922810

[1713922810]

1713922810

[DaveF]

Hey, just want to plonk a link to this topic here for anyone interested in running a LN node in the USA:
Devastating "Infrastructure" bill in US - contact your representatives

Apparently they want to treat node operators as brokers due to 'facilitating transactions'.
Maybe worth following if you're located there..

Kohas787 will be so happy that he can start spam-calling the nearest FBI office, and report that his neighbor is running a Lightning node from the basement, earning $5.00 in routing fees per week.

If only somene wrote a guide in how to run a Lightning node over the TOR network, https://stopanddecrypt.medium.com/running-bitcoin-lightning-nodes-over-the-tor-network-2021-edition-489180297d5

And for those who are using a 'node in a box'

If you are using a Raspiblitz to run over tor it's s single click and a reboot.
For mynode it's insert $99 and click a and reboot as it's a premium feature
Umbrel also can use TOR, not sure of how to configure it. Setting up an umbel box is on the to-do list.

So if you are a bit twitchy about setting up / running a linux box you can still run it all on TOR thanks to the work of the people who made these devices.

-Dave

[darkv0rt3x]

...Let me tell you about another great solution. And if you're willing to spend a bit more and use an NVME M.2 drive, consider a RockPro64 SBC. It's way faster than Raspi4 and it supports the mentioned superfast drives....

Did you find / do you have a nice case that supports that card?
Everything I have seen is either big and expensive or looks like cheap plastic.

I did my own. It's not ready yet because something got in the way in the meantime. But I bought a board of acrylic, took some measures (quite some trial and error, tbh, :p) and cut it, glued it with specific glue for acrylic, used some plastic bolts and threads to place the board on top of them, and I'm now (this is the part that is stalled due to some life issues) trying to program a small chip to control the NVME M.2 drive temperature, conenct the fan to it and then screw the fun as close as possible to the NVME drive chips. It will also use a thermal resistor to read temps and the fan will change rpms accordingly!

The ones I saw (cases) in the official site are too big, metal and expensive, for my taste. And don't even remember if I could mod them to reuse the old fans I had here!

[n0nce]

Trying to program a small chip to control the NVME M.2 drive temperature, conenct the fan to it and then screw the fun as close as possible to the NVME drive chips. It will also use a thermal resistor to read temps and the fan will change rpms accordingly!

Wait, since when is SSD temperature such a big concern? And since when does a LN node have so much disk I/O going on that it would get super hot?
I run 2 NVMe SSDs in a tower of mine with just simple passive heatsinks on them and they run just fine.
Just a heads up that you might be overengineering here!

[DaveF]

Trying to program a small chip to control the NVME M.2 drive temperature, conenct the fan to it and then screw the fun as close as possible to the NVME drive chips. It will also use a thermal resistor to read temps and the fan will change rpms accordingly!

Wait, since when is SSD temperature such a big concern? And since when does a LN node have so much disk I/O going on that it would get super hot?
I run 2 NVMe SSDs in a tower of mine with just simple passive heatsinks on them and they run just fine.
Just a heads up that you might be overengineering here!

These are not PC towers or even micro towers. These are very tiny cases with a RPi4 (or rock 64 in this case) and the m.2 drive. There is almost no airflow so you do have to make sure there is enough to keep everything happy. Even a couple of watts in a tight space can get very hot very quickly if there is no air movement.

-Dave

[darkv0rt3x]

Trying to program a small chip to control the NVME M.2 drive temperature, conenct the fan to it and then screw the fun as close as possible to the NVME drive chips. It will also use a thermal resistor to read temps and the fan will change rpms accordingly!

Wait, since when is SSD temperature such a big concern? And since when does a LN node have so much disk I/O going on that it would get super hot?
I run 2 NVMe SSDs in a tower of mine with just simple passive heatsinks on them and they run just fine.
Just a heads up that you might be overengineering here!

You're right. I forgot to mention. The temps goes higher only during IBD, when I eventually have to do it. Other than that, it runs smoothly with no major temperature issues.

Trying to program a small chip to control the NVME M.2 drive temperature, conenct the fan to it and then screw the fun as close as possible to the NVME drive chips. It will also use a thermal resistor to read temps and the fan will change rpms accordingly!

Wait, since when is SSD temperature such a big concern? And since when does a LN node have so much disk I/O going on that it would get super hot?
I run 2 NVMe SSDs in a tower of mine with just simple passive heatsinks on them and they run just fine.
Just a heads up that you might be overengineering here!

These are not PC towers or even micro towers. These are very tiny cases with a RPi4 (or rock 64 in this case) and the m.2 drive. There is almost no airflow so you do have to make sure there is enough to keep everything happy. Even a couple of watts in a tight space can get very hot very quickly if there is no air movement.

-Dave

Yeah that is a fact too. My RockPro64 chip is running a medium heat sink from Pine store, a small PC RAM fan (Revoltec brand). And it probably didn't even need, but it's always good. The NVME drive don't even have a heatsink, so, there's where my concern with temps come.

[n0nce]

These are not PC towers or even micro towers. These are very tiny cases with a RPi4 (or rock 64 in this case) and the m.2 drive. There is almost no airflow so you do have to make sure there is enough to keep everything happy. Even a couple of watts in a tight space can get very hot very quickly if there is no air movement.

That's good to keep in mind! I haven't tried yet with a good passive heatsink and no airflow, to be honest, maybe that's indeed not enough.

[DaveF]

But i wonder how reliable is the "one-click Tor" feature? Would it work automatically if Tor is also censored? It's not easy task to find bridge (which isn't blocked) or create your own bridge.

At least for raspiblitz there are no 'quick' options to change. If TOR is censored by your ISP / government you would have to change the settings manually in the TOR configs.

I can't check on mynode at the moment since I reformatted it to try Umbrel yesterday and had a few issues so I stopped trying. I'll try again next weekend unless I have some slow time during the week.

-Dave

[DaveF]

One more question, if you check Tor config, are there any warning that you shouldn't edit it by yourself or the change might be overwritten by raspiblitz? I know few Tor tools which suddenly change your configuration which might broke other tool.

If you were on clearnet it goes and changes the bitcoin.conf and all the other apps you have installed to use TOR.
If you un-check it, it puts the configs back to the way they were.
If you check it again, it's back to the stock TOR config.

As far I can tell with over a year of use it never touches the configs again. The configs are not even on the SD card but on the HDD, so when you upgrade they are the way you left them.

But you are probably better off asking on their github for more info: https://github.com/rootzoll/raspiblitz/issues

-Dave

[darkv0rt3x]

Hello.

What are your thoughts about triangles and other geometries between LN channels?

I have the idea that these kind of closed loops of channels are bad for the network, because it's like if you are putting the entire network out of that loop. So, wouldn't it be more reasonable to spread the most possible our connections instead of creating those closed loops between channels?

[Syke]

Hello.

What are your thoughts about triangles and other geometries between LN channels?

I have the idea that these kind of closed loops of channels are bad for the network, because it's like if you are putting the entire network out of that loop. So, wouldn't it be more reasonable to spread the most possible our connections instead of creating those closed loops between channels?

A newly-opened channel only has outbound liquidity. Loops give all nodes both inbound and outbound liquidity without creating redundant channels. It might even assist in rebalancing.

[darkv0rt3x]

Hello.

What are your thoughts about triangles and other geometries between LN channels?

I have the idea that these kind of closed loops of channels are bad for the network, because it's like if you are putting the entire network out of that loop. So, wouldn't it be more reasonable to spread the most possible our connections instead of creating those closed loops between channels?

A newly-opened channel only has outbound liquidity. Loops give all nodes both inbound and outbound liquidity without creating redundant channels. It might even assist in rebalancing.

Not 100% true now.

(I speak for myself) - With C-lightning you can open a channel that is balanced right from the start if you use liquidity ads, which were recently introduced.

They give you inbound and outbound capacity but in separate channels. I think it's always better to have balanced channels. I mean, it's better to have good balance per channel than just globally, which is what you might get from rings!

And what you mean by "redundant channels"??

[Syke]

They give you inbound and outbound capacity but in separate channels.

Yeah, that's redundant. It is better to have 1 big channel than 2 redundant small channels.

I mean, it's better to have good balance per channel than just globally, which is what you might get from rings!

True, rings start out unbalanced and need manual balancing. Dual-funded channels would be ideal.

[n0nce]

Hello.

What are your thoughts about triangles and other geometries between LN channels?

I have the idea that these kind of closed loops of channels are bad for the network, because it's like if you are putting the entire network out of that loop. So, wouldn't it be more reasonable to spread the most possible our connections instead of creating those closed loops between channels?

Sure, if you only have one channel and that channel is part of a LN ring, the rest of the network is out of this loop.

But nobody does this. As soon as you have multiple of these rings, everyone in one of these 2 rings can reach everyone in the other ring, since you run the node that ties them together. Of course this gets even better as soon as every participant has multiple rings or triangles and connections to hubs and other participants.

In general, a mix of hub, ring, triangle, individual connections is what I personally use

It's true that when creating rings the channels are single-funded and your 'balanced liquidity' is only 'globally'. It shouldn't be an issue with low per-hop fees, but if you want, you can balance a ring once it's completed.

[ndalliard]

It's true that when creating rings the channels are single-funded and your 'balanced liquidity' is only 'globally'. It shouldn't be an issue with low per-hop fees, but if you want, you can balance a ring once it's completed.

i did once a triangle and we set the fees to zero after every channel was open an balanced the triangle so that everyone had a 50/50 channel. after that we set our fees at whatever we wanted

[n0nce]

It's true that when creating rings the channels are single-funded and your 'balanced liquidity' is only 'globally'. It shouldn't be an issue with low per-hop fees, but if you want, you can balance a ring once it's completed.

i did once a triangle and we set the fees to zero after every channel was open an balanced the triangle so that everyone had a 50/50 channel. after that we set our fees at whatever we wanted

Yup, I briefly mentioned this possibility:

if you want, you can balance a ring once it's completed.

[darkv0rt3x]

Oh, ok, good idea, actually. To set the fees to 0 and then, balance the channels and lastly set the fees again!

I'll see if my ring can/want to do that. I just created a ring with 2 other people.



Anyways, for the ones interested, Andreas M. Antonopoulos (@aantonop), Olaoluwa Osuntokun (@roasbeef), Rene Pickhardt (@renepickhardt), colaborated in another great work for the Lightning Network community. An online book. Check it here:

https://github.com/lnbook/lnbook

Hope this is useful to as many of us as possible!
Thanks to their great work.

[psycodad]

I am posting this reply here, it was Rath_'s reply to my post in the bitcoind, c-lightning and RTL on centos 7 (walktrough) thread and he rightly suggested to come here with my questions.

If anybody has any reading pointers on how fees work in lightning, what good settings are etc.

There are no good settings. If you open a channel between two large nodes then you will probably not route any payments unless you set your fees to zero. You have to experiment with your fee settings. I opened a large channel to Bitfinex and a small one to Nicehash. Even though my fees in that channel (Bitfinex) were high (1 sat base fee; 95 ppm), I routed quite a few payments. The largest one was ~950k satoshi and I earned ~91.5 sat in routing fees just for that single transaction. After some time I had to lower my fees since no one wanted to send their payment through that channel.

Are you familiar with the way the fees are calculated or do you need an explanation?

No, to be honest I do not seem to understand the fee system fully yet. I have a solid understanding of how the fees work on-chain but not in lightning. From what I've seen there is a base fee and a percentage of the amount to be sent?

Also, I would be interested what the settings of "urgent", "normal", "slow" etc. mean when setting up a new channel, I guess urgent translates to high fees, but I am not sure yet.

Honestly, I would not use RTL to open and close channels. I overpaid a few times because of it. Also, you can open multiple channels in a single transaction via a command line! This way you can save a ton on money on the transaction fees.

Good point, I guess that's because RTL sets the on-chain fee for the funding transaction high enough to get into the next block at anytime and by doing it manually you can use 1-2sat/byte and just wait some time.

Are people setting their fees on a per-channel basis or globally?

Most people set their fees on a per-channel basis. If you are running a small node then you will very likely change them often. There is a plugin which automatically adjusts the fees for each channel but I am not sure how well it works.

Oh plugins, that's one open construction site I still have  Didn't really work out yet, there are a few I would like to use, like the sqlite3 duplication and ev. autopilot but I can't get them to run yet. But I understand the general idea is not to care about the defaults, but fine-tune each channel so that it starts routing payments.
What are some ballpark figures for like dirt-cheap but not free and what are some upper bounds before it becomes criminal price gouging (yeah I understand, the network just routes around expensive nodes, but you should get the idea what I try to ask for)?

By the way, are you interested in opening a dual-funded channel at some point? It is still an experimental feature but I have successfully opened this type of channel with two other bitcointalk members. Here's my node for reference. Also, you might find The Lightning Network FAQ useful. Most of the LN related discussion is held there.

I am all in, just need some time to get more experience in operation of my node before I waste your time. I will happily get back to you by PM in the next days. I still need to finalize the setup in terms of DR plan / backup / resilience. I'll still need some assistance in creating this via rpc interface.

Thanks for the advice to move my questions to here, which I did.
I should have posted my questions here in the first place as you suggested, I only wanted to let mocacinno know how much his walkthrough helped me.

[Rath_]

No, to be honest I do not seem to understand the fee system fully yet. I have a solid understanding of how the fees work on-chain but not in lightning. From what I've seen there is a base fee and a percentage of the amount to be sent?

Yes, you can set two parameters for each channel: base fee and fee rate (also known as PPM - pay-per-millionth). The fee formula looks like this:

base_fee + (forwarded_amount * fee_rate/1000000)

When you route a payment, you receive some coins in one channel and you spend coins from your other channel. You don't charge someone twice for your service. The fee paid by the payer is calculated based on the fee settings of the channel which forwards the payment to the next node.

Didn't really work out yet, there are a few I would like to use, like the sqlite3 duplication and ev. autopilot but I can't get them to run yet

I am not sure how well c-lightning's autopilot works. I believe that LND's autopilot opens channels mostly to large nodes. That's not really what you want if you want to maximize your routing earnings. Looking for good peers can take a lot of time, but it's kind of fun.

What are some ballpark figures for like dirt-cheap but not free and what are some upper bounds before it becomes criminal price gouging (yeah I understand, the network just routes around expensive nodes, but you should get the idea what I try to ask for)?

That's a tough question. I wouldn't open a channel to someone who charges more than 100 PPM in most of their channels.

[psycodad]

Thank you for your reply and explanations so far.

Yes, you can set two parameters for each channel: base fee and fee rate (also known as PPM - pay-per-millionth). The fee formula looks like this:

base_fee + (forwarded_amount * fee_rate/1000000)

When you route a payment, you receive some coins in one channel and you spend coins from your other channel. You don't charge someone twice for your service. The fee paid by the payer is calculated based on the fee settings of the channel which forwards the payment to the next node.

Thanks a lot for that, I think I got it now how fees are set.
What I am still unclear is how the fee collection works, are you saying that if Alice pays to Bob over nodes X, Y and Z, X charges a fee to Alice and pays from that Y fees, who in turn pays Z fees?
Like in:
Alice - [X] - [Y] - [Z] - Bob
So fees have to be X^fee > Y^fee > Z^fee, otherwise somebody is going to make a loss?
Or if I am for example X, would I calculate Y's fees and add it ontop of mine and charge the sum to Alice, like in calculating total cost of a specific route?

Sorry for the stupid question, I just read https://medium.com/suredbits/lightning-101-lightning-network-fees-86abbbc17024 and I am none the wiser still.
EDIT: Just found this one, and it says Alice pays the sum of all fees to X, X pays Y and Z fees to Y etc.

I am not sure how well c-lightning's autopilot works. I believe that LND's autopilot opens channels mostly to large nodes. That's not really what you want if you want to maximize your routing earnings. Looking for good peers can take a lot of time, but it's kind of fun.

Yeah I started to figure that negotiating channels with peers is going to be the fun part, almost like negotiating peerings in the old internet days. I was interested in the autopilot to a) get a grip on how plugins work in c-lightning and b) to see what it would do in dry-run. But I am all game for the social fun part of establishing channels  

What are some ballpark figures for like dirt-cheap but not free and what are some upper bounds before it becomes criminal price gouging (yeah I understand, the network just routes around expensive nodes, but you should get the idea what I try to ask for)?

That's a tough question. I wouldn't open a channel to someone who charges more than 100 PPM in most of their channels.

That's something to start off from, thanks a lot.
And for the base fee, free/0?


Also I am wondering about tor vs. clear-net nodes, is it possible to "dual-host" a node on tor AND clear-net at the same time or do I have to stick with one or the other?