THE CHIA FARM

[Rand__]

Then the second problem is the flat network.
During plots transfer, the system is not fast enough to solve Chia challenger in 30 seconds.
Big mistake, when I moved plots to plotter to harvester, the main network is saturated.

Solution:
I deployed a private network to transfer plots between harvester and plotter.

Just noticed I had that issue too, always thought it was disks becoming saturated but it turns out it was actually network.
Not that i did not start with 10G+ connections, but became sloppy running some dirty fix or the other...

Stopping 2 transfers brought proof time from >5s to 0.5s-1.5s again

 

[gb00s]

These are 8min of proof times while copying 3 plots from temp dest do final dest for my 2nd
EDIT: Seeing the 1st log for '3 plots were eligible for ....'

ADD: Adding 2x plots to the storage and nothing changes regarding proof/response times.

 

[msg7086]

Tip about saturating network and hard drive -- always move the files manually, and set -d and -t to the same place. With moving files manually, you can run rsync with bwlimit to reduce the impact to network and hard drive. You may also want to set ionice to realtime for harvester, and idle for moving process.

 

[bash]

Just remember ionice may not work on your system for example ubuntu uses deadline now which does not support priority. I hope someone can prove me wrong.

 

[gb00s]

Just remember ionice may not work on your system for example ubuntu uses deadline now which does not support priority. I hope someone can prove me wrong.

Deadline is read-oriented. Worst case your write request is unanswered for 3sec, expires and has to jump back in the queue with a new request. What about mq-deadline with multi-queuing?

EDIT: Just found this ...

Please note that looking up qualities for plots passing the filter requires about 7 random reads in a plot, whereas actually looking up a proof requires 64 reads. It might not be feasible on a slow NAS, since these are sequential reads. Furthermore, you need to take into account network latency to propagate your proof and block to the network, so you should be under 5 seconds to reduce risk of losing rewards.

 

[msg7086]

Just remember ionice may not work on your system for example ubuntu uses deadline now which does not support priority. I hope someone can prove me wrong.

Thanks, so it seems I'm already using mq-deadline so ionice doesn't do anything to me I guess? Good to know.

 

[funkywizard]

Yeah, easy to forget, but while its friendly rivalry for most of us here (I'd assume), its a tough business worth a lot of money for some other folks out there :/

The minimum transaction fee is zero. I'm astonished that nobody has ddos'ed the mempool with tiny transactions.

 

[ari2asem]

i see lots of comments about syncing node, chainblock heigth (locally and other heigths) etc etc...2-3 days ago i asked the question whether it is possible to add address's of the nodes of your choise in config.yaml.
maybe this way we could exclude nodes that are "out of sync".

so my question....can we manualy add our own choise of address (=nodes) ??

 

[funkywizard]

i see lots of comments about syncing node, chainblock heigth (locally and other heigths) etc etc...2-3 days ago i asked the question whether it is possible to add address's of the nodes of your choise in config.yaml.
maybe this way we could exclude nodes that are "out of sync".

so my question....can we manualy add our own choise of address (=nodes) ??

It might require a custom patch (publicly available but not mainline version), but yeah we do that.

 

[ari2asem]

It might require a custom patch (publicly available but not mainline version), but yeah we do that.

you do that ?? or you shall do (in the future) ??

 

[Rand__]

It might require a custom patch (publicly available but not mainline version), but yeah we do that.

Can't you add nodes of your choice already via the gui? Then i'd assume there should be an interface for this or a place the temporary data is saved that could be modified via script.

O/c limiting only to "approved" nodes would require regular verification and potentially restarts if you can't delete on the fly, but should be possible

 

[ari2asem]

couple of (stupid or basic) questions for Windows and chia GUI:

- how do you run plotting machine and harvesting machine ? are those 2 different computers in 1 network, with UPNP set to FALSE inside config.yaml for both computers ??

- if the answer to my first question is YES, then you run CHIA.exe on these 2 computers at the same time ?

- in config.yaml i can see my farming/harvesting directory.
can i manualy add (not via GUI) inside config.yaml as many directories (on different hdd's in 1 computer) as i want ??

- for harvesting/farming computer do i need fast internet connection (from my modem/router to the ISP) ??

- how many times i need to run command PLOTS CHECK N-30 to increase my lottery chance's ? or this does not matter at all ?

- how can i check whether my computer is responding within 30 seconds on requests from chia network ??

thanks in advance ....

 

[Rand__]

couple of (stupid or basic) questions for Windows and chia GUI:

- how do you run plotting machine and harvesting machine ? are those 2 different computers in 1 network, with UPNP set to FALSE inside config.yaml for both computers ??

You can plot and harvest/Farm on the same or different boxes.
If your setup is small enough you can run two full nodes, I simlpy told windows to turn of device discovery=upnp, this worked reasonably well (5 full nodes) up until 600 plots

- if the answer to my first question is YES, then you run CHIA.exe on these 2 computers at the same time ?

Yes

- in config.yaml i can see my farming/harvesting directory.
can i manualy add (not via GUI) inside config.yaml as many directories (on different hdd's in 1 computer) as i want ??

I think so, but remember python is picky about layout (tabs, spaces etc)

- for harvesting/farming computer do i need fast internet connection (from my modem/router to the ISP) ??

Harvester no, Farmer yes.

- how many times i need to run command PLOTS CHECK N-30 to increase my lottery chance's ? or this does not matter at all ?

No need to run n 30, its just to ensure the plots are ok, n5 1/day should be fine

- how can i check whether my computer is responding within 30 seconds on requests from chia network ??

This has now being bumped to WARNING level, so running a Powershell Tail on the Log grepping for WARNING should suffice

 

[msg7086]

Plotting can be done offline, all you need is public key and chia binaries.

You should only need one full node. Pick a computer that is relatively powerful (2 fast cores or above) and stays online 24x7, and rest of your network can run harvester only. It's also a wise choice to run full node on a 2+ core VPS with good connections, and run harvesters at home. A harvester reconnects in seconds, but if a full node disconnects it may take a few minutes or sometimes hours to reconnect.

plots check is not necessary at all. It only checks if it's good or invalid. Unless you have a failing disk to store plots, you don't need to check them at all, ever.

 

[Bert]

Ok update on my side, got another 2 last night with 1900 plots. Now, I may have another shot to get 2 more before netspace is hitting 50exabyte and that would cover my expenses if I can sell all of the coins with today's price.

I am not sure what to do now, I have several SMR/USB 8TB and 5TB drives I bought from Costco with the cost of $16.5 and $20 per TB. I refrained shucking them as I wasn't sure if I will continue. Transfer speeds to these drives are horrible at 35MB/sec so I am pretty much bottlenecked behind this speed hence I am not adding more space.

I have an order of 16TB USB Seagate's coming at $20 per TB, I was told that they will be getting fulfilled next week. Pre chia prices on these drives looks pretty good to me.

I have plenty of disk shelve space available and I think I can generate 70-100 plots a day. Hence I can continue efficiently if I want to as long as I buy hard drives.

The math to make decision here is a little bit complicated because I am not sure if I will ever get my money back on the hard drives as it is hard to predict netspace growth. My guess is that netspace growth will accelerate as the tooling and expertise become wider. There is also possibility of hard disk crash if chia falls apart but that seems unlikely to me unless company starts to disperse their 21M coins.

My options are:

1. Cancel/return all the hard drives I purchased.
2. Keep the hard drives and try to scalp them
3. Continue to expand until netspace hits 50Exabyte and hope for the best.
4. Sell all my hard drives and gear while the prices are peaked. I know I have much more gear than I will ever 'need" unless I do chia mining.

What would you do?

 

[msg7086]

The ROI is definitely a lot worse than early days. But if you get a few coins and sell them your cost on rigs would probably be covered, so the rest of the days are basically no cost farming. I would continue with what you have.

That said, if you decide to stop I'd happy to take all your 16TBs at $20 per TB or so. (I must be dreaming, right?)

 

[Rand__]

Congrats (and f.u. as it is customary on the chia reddit)

You could get rid of the SMR drives for same or more cash (or return) and only use the good drives.
Its still early in the Chia lifecycle so no reason to quit now, with pools coming in a month (or 3) you still can quit if the steady income exceeds your running costs. Scaling up to match pool growth is mandatory only for solo plotting i think - o/c youll get less rewards with pools too, but then it will be way easier to decide if its worth it - now its only luck whether you get ROI or not.

 

[Rand__]

That said, if you decide to stop I'd happy to take all your 16TBs at $20 per TB or so. (I must be dreaming, right?)

You're dreaming He promised them all to me.

 

[Bert]

That said, if you decide to stop I'd happy to take all your 16TBs at $20 per TB or so. (I must be dreaming, right?)

May be not. I will let you know.


Yeah another problem is selling the coins, it is kind of scary to send thousands of dollars to some chinese exchange. Do we know why coinbase and others are not opening up chia? That would also push the price higher.

 

[gb00s]

.... Do we know why coinbase and others are not opening up chia? That would also push the price higher.

... quite the opposite. Why are so much XCH been hold? Yeah, to finance the project ... sure, if everybody can sell their Chia. ... easily