Interesting that big nodes like bitrefill don't have real redundancy.
https://allesvoorbitcoin.substack.com/p/lightning-network-our-high-maintenance
8 years of Lightning Network : our high-maintenance crazy ex
Discussion
If it can happen to coinos, then it can happen to anyone of them I guess.
Yeah, interesting. I guess we do not know how different companies run their infrastructure.
I have been reading a little bit on the topic. Seems like it is possible to run a highly-available lightning node, using kubernetes for example:
https://github.com/lightningnetwork/lnd/discussions/8040
I guess it should also be possible to run a highly available postgres DB for the lnd instances, also using kubernetes.
That would solve a few tech issues, but not all.
Kubernetes will cause incosistent channel states.
The core "feature" (or flaw, whatever you'd call it) is that LN is inherently secure and has a secure handoff to not have anyone cheat and duplicate channels and the funds therein.
Without state synchronization when one node or datacenter location goes belly up, one active node instance can safely manage channels at a time. Running multiple active replicas risks channel inconsistencies, as noted in discussions about LND’s database synchronization problems.
Are those channel states stored in the DB, or in the LND instance itself?
In the case of the database, if the database itself could be hosted in a highly-available way with fail over, etc, then that issue would be taken care of.
Have you seen this repo by any chance? https://github.com/Filiprogrammer/lnd-ha-guide
(I've been reading some more...) :)
Would love to hear the take of someone who does this professionally.