off hours - for me, could be when I leave “the office” and I can shot down a test-server (which would run a vm with chia).
Thanks for your input. I get it. But, I’ve experienced that sync is relatively slow (12 hours off takes 3-5 hours to sync), and I’m concerned about having 2 full nodes on the same IP.
I have an IPV6 prefix, and can assign unique public addresses, but I see only few IPV6 connections to other nodes.
Is there a problem with running two full nodes on the same LAN?
I mean, as long as you disable on one UPnP, and not try to somehow port-forward 8444 to that other node, it should run fine, and not disrupt the primary one. My take is that you should be able to run as many nodes as you want.
Can we say same network each node with different IP, syncing both nodes would just take up extra bandwidth and slow down the process. I guess that why we would have a main node farmer and 2 harvesters instead of three farmers.
Syncing bandwidth is basically zilch. Even when you sync from scratch, you rarely go over 2Mbps (250KBps), so you have plenty of nodes on 100Mbps downstream speed.
Nah. The problem is only with peers that want to connect to your nodes. You can have one node (main) that will have UPnP active (no need for port-forward, unless your router blocks UPnP), and the rest of your nodes with disabled UPnP, so no peer would connect to them, but all of them will be able to make connections. No problem at all, and really no concerns about bandwidth (yeah, those syncs from scratch going on for 2 days or more are exactly the effects of the bandwidth being so tiny - you can use your old analog modem, OK, maybe your DSL one).
You can look at that as FROM and TO addresses. The TO address needs to be 8444, but FROM will be automatically selected by your system and then your router (therefore you can have more than one peer when reaching out). So, this is the reason why only one node can have inbound connecting peers. However, all peers are free to connect to any peers you want.
Your node needs quite a bit CPU power, and a fast drive (for db) not really fast bandwidth, so when you just run a harvester, you can run it on a tiny box. Actually, just for testing, I am running FlexFarmer (basically a harvester) on RPi 2 (it has 1GB RAM). Although, the CPU bandwidth (to peripherals) is kind of slow, so proof times are not that great.
But for a $100K, most likely you could buy a washing machine with one 15" platter. The good part is that IBM was recommending pure alcohol to clean it.
I am just waiting for them to start shipping (RPi 3 is perfect for Chia harvester, actually on CM4 module you can add PCIe based SATA card - SAS controller!!). You cannot buy anything, unless you pay through the nose to scalpers.
Actually, the problem with that port-forwarding for 8444 is related to chia DNS introducer. It should collect not just an IP, but also the port you want to dedicate to it. This way, you could port forward different ports to different nodes. As it is done right now, the port is fixed, thus all those problems.