The two major nostr relays are not active in development anymore.
This is ok?
Discussion
It's ok, they made the perfect code just like Satoshi did
You never run a relay huh?
Bro I run like 60 relays
And you dont realize how poorly they are built?
How much traffic your mastodon server? Just curious. Ya, I'd say strfry is the most well built relay you can imagine.
You haven't tried Nosflare yet 😤😂
Does it do reqs yet?
Strfry is not good at all. Its not built to scale and lmbd is trash for this use case. Its way too disk heavy . Development stopped last year.
M, no it serves the traffic direct from memory mapped storage
Lmdb uses mostly ram
You didn't answer my question, how much traffic? 🦜
Sorry, Nostrudel keeps hiding that post .
One Active instance I self-host uses 1TB of data per month. When my instance had a few users, it was under 1TB I think. CF combined them all so it's hard to tell in the monthly report.
Well I guess it's hard to determine then cause I only serve 5tb of pure text for 60 relays
That seems like a lot. Are they small relays? Do they mirror large ones?
They are organic relays
What does that mean?
It means that I'm not attempting to network with other relays. Whatever happens happens
Wow interesting. Nostr seems like it's worse than the Fediverse software. Having clients do all the server-side stuff means more traffic.
And my load avg is only 1-2
Why was lmdb deprecated?
Always has 🤙
Can you move from CF to on-prem?
It's CF specific atm. But I am looking at possibly releasing a platform agnostic version. Though, might maintain core focus with only CF so as not to make it too fragmented.
Last I looked it just did individual events... 🤔
Ah yes, very first iteration, the filtering had a bug. Was fixed shortly after. Now it does everything a normal relay can do, except constant streaming of new events. But working on that. Currently, it closes the connection after each REQ.
Scope it: https://github.com/Spl0itable/nosflare
I also am making a point with my OSS to have zero cloud lock in. It's just haproxy, and strfry. Maximum self hosting
i'm releasing a new one 27 may called replicatr
it has a storage management cache capability and can be interfaced with a second level (distributed, possibly immutable) data store
it already can do live subscriptions as it's based on khatru, but i have massively rewritten the nostr library, the relay logic and the event store
it uses badger, and i am right now in the middle of completing a revision of the garbage collector which currently can measure utilization of the store in 8 seconds for a 14Gb of event data and should be viable for up to a terabyte with some tuning (later i plan to make a self-tuning GC timer with dynamic metadata calculation)
https://github.com/mikedilger/chorus is better