I’m not sure about the math, just putting it out here so people can correct it and we get consensus fast.
Total nbr of plots in netspace: 165.212.160
Your plots: 4500
Your chance of having the best proof for a challenge (winning): 4.500 / 165.212.160 = 2,7237704537002603198214949795463e-5
The rest of the netspaces chance of winning (you losing) : 1-2,7237704537002603198214949795463e-5 = 0,9999727622954629973968017850502
Your chance of losing every challenge 20 days in a row: (0,9999727622954629973968017850502)^(4608*20) = 0,04917778683189503804403832688622
Ah congrats! Can I ask was the proof found on one of the harvesters or your farmer?
I too am worried if there is something inherently wrong with the harvester setup, I only have about half of your plots mind but have seen no wins since moving my setup over to using harvesters.
I’m experiencing the same phenomena. I was winning Chias round about the expected time until I expanded and created a few other harvesters. It has already been 2, almost 3 times longer than expected time and nothing. It really is disheartening, seems like the odds of winning is higher when everything is in one machine.
Wouldn’t say this proves that that architecture is superior… if that’s the case Chia devs have made a HUGE mistake. I suspect the proof is just a coincidence. If not, get that bug report in the git repo.
As I had problems of a too high latency I moved to a multiple harvester model as well and didn’t find any blocks since then. I have my main wallet harvest and farm and have another harvester connecting to the main wallet. Currently I “should” find one every 10 days or so, but no blocks since 18 days now. I do see my harvester and main farm plots pass the filter so I guess it’s just bad luck, but I will keep you posted.
Maybe a bit offtopic but how is the estimated farm size calculated on this website?
For me it’s showing 80tb while it’s actually 350tb. Should I be worried?
As an anecdote for ya… I only found blocks after going to multiple harvesters. I won 6 blocks in May, started with 3 harvesters, ended up with 5 by the end of the month… over 2700 plots now.
Would love to have you and OP compare notes about the configuration of your multiple harvesters. AFAIK you just swap a config on the harvesters to use the farming node value, import the CA keys, disable UPNP, and it’s all good.
With that many plots sometimes 10+ plots passes the filter. They all have to be looked up over the network. If there is also some plots being written over the network it’s way too slow. (30 seconds +)
I had a few solutions for it:
plot offline to an USB disk and move the plot files manually every now and then to reduce the network load.
upgrade my gigabit network
let the machine with all the disks harvest on the same pc.
I chose option 3. I do see this works beautifully in term of lookup latency, but just didn’t find anything yet. I know how it works therefore my “should” was between quotes already I guess I’m just less lucky this round, but I will keep an eye on it as it should even out over time.
I found no difference in the effected plot locations on a gigabit ethernet vs local usb, if there was high IO load like, when running check plots the challenges times would go way over the 30 second threshold. Otherwise they are at worst around 6 seconds with about 10 blocks passing. Pretty weird and undercommunicated limit (needs awareness of log files, not exposed via GUI) for single harvesters, that also voids the whole green, “can run on rasperry pi” angle.
upnp is not a Windows protocol, it is independent. More specifically you should look for the " Internet Gateway Device Protocol" part. Whether or not it is installed in your distribution is another thing.
I guess I’m just less lucky this round, but I will keep an eye on it as it should even out over time.