THE CHIA FARM
- Thread starter gb00s
- Start date
Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.
As embarrassing as this might sound, but I have the impression I got totally confused about what a node, harvester, farmer or plotter is.
I have 'Server A' where I plot, and store finished plots together. Server A is full now. Can only plot and move the finished plots to 'Server B'. On 'Server B' I'm also plotting and storing now newly finished plots. 'Server C/D/E' are plotting locally only and transferring their finished plots to 'Server B.'
All 5 servers have the same keys installed. All are started as 'chia start node farmer wallet'. Reading your comments, this sounds utterly wrong to me now. So who is who ...
I have 'Server A' where I plot, and store finished plots together. Server A is full now. Can only plot and move the finished plots to 'Server B'. On 'Server B' I'm also plotting and storing now newly finished plots. 'Server C/D/E' are plotting locally only and transferring their finished plots to 'Server B.'
All 5 servers have the same keys installed. All are started as 'chia start node farmer wallet'. Reading your comments, this sounds utterly wrong to me now. So who is who ...
Let me describe my setup. current and future
Full node , with private key on this machine, also act as farmer, open to the internet with port forward 8444
Current, I oversize the machine, E5-1275v3 , 32gb , my first encounter with Chia, using Window GUI version,
Future , move the full node to a HP 800 G1 mini, low end I5 or I3.
May be a linux VM , may move from Window to Linux.
Harvester, each pod is a Low power ITX version , very low end CPU ( Celeron ) , 8gb ram, holds 4-8 drives.
Try to keep power usage as low as possible . not extra 10gb or 40gb network card,
extra card consumes 5-10w and is not needed.
currently, I have 5 harvester
May combined to a chassis that house 10 drives, still keeping the low end CPU such as ( G4560, G4900, G5400).
Consolation will depend how long I'll continue farming.
Plotter, 16Core/32threads , 128gb ram , power hungry monster. lots of SSD /nvme .
Power up to plot only. shut off when it is not in use.
Full node , with private key on this machine, also act as farmer, open to the internet with port forward 8444
Current, I oversize the machine, E5-1275v3 , 32gb , my first encounter with Chia, using Window GUI version,
Future , move the full node to a HP 800 G1 mini, low end I5 or I3.
May be a linux VM , may move from Window to Linux.
Harvester, each pod is a Low power ITX version , very low end CPU ( Celeron ) , 8gb ram, holds 4-8 drives.
Try to keep power usage as low as possible . not extra 10gb or 40gb network card,
extra card consumes 5-10w and is not needed.
currently, I have 5 harvester
May combined to a chassis that house 10 drives, still keeping the low end CPU such as ( G4560, G4900, G5400).
Consolation will depend how long I'll continue farming.
Plotter, 16Core/32threads , 128gb ram , power hungry monster. lots of SSD /nvme .
Power up to plot only. shut off when it is not in use.
4 "big machine" 16c/32t @ 35 to 38 plot each = 4 x 35 = 140 per day, I got the E5-4667v3 $85 each long time ago.
4 x 4c/8t e5-1240v3 @ 12 plots each = 48 plots per day
180 plots day.
I have more CPU to build few more big machine.
Limitation is HDD space, I am committed 200TB disk space to farming.
Unless I won another 2 XCH, then I would commit another 200TB disk space.
I am about 3 days to fill up 200TB.
4 x 4c/8t e5-1240v3 @ 12 plots each = 48 plots per day
180 plots day.
I have more CPU to build few more big machine.
Limitation is HDD space, I am committed 200TB disk space to farming.
Unless I won another 2 XCH, then I would commit another 200TB disk space.
I am about 3 days to fill up 200TB.
C, D and E should be plotters, they don't need the private key. You can stop the chia and run the plotters from commandline.
A is harvester/farmer but it should have been only harvester. You don't need the private key over there. You need to start chia in harvester mode.
B could be your farmer &plotter. Here you can have your private key, start chia full node. Your harvester can talk to you the farmer to serve challenges. This is the recommended set up by Chia.
This setup looks awesome. I need to also separate my plotter from the farmer to save power but my network is still 1G so couldn't do that. Marsh must have a great home lab.
Marsh, How many plots can your plotter generate in a day?
Ah awesome deal on E5-4667v3 . Very nice set up. Why do you wait for the 200TB? If you have the space, power cost of plotting is insignificant in comparison to possible return.
Depend on how many machines.
You could tried some USB 2.5gbe or PCIe 2.5gbe , direct connect , no need for a 2.5gbe switch.
Amazon sell the USB 2.5gb ethernet very cheap ( $19 each )
PCie card on Ebay is about $22 each when buy two.
I picked up a 8 port 2.5gbe trendent switch for $140,
( add to the cart $150, wait till the next day to check out, price became $140 )
I don't know how to make direct connect. Is this the one: Direct_Connect - Debian Wiki
My farmers and plotters has 10gb nics but I don't have the switch( I have the incomplete switch I built with your guidancd) I might need to get your expert advice again.
You know I have the supermicro board with 6 10GB NIC, that would be an awesome harvester/farmer head if I can make this direct connect work.Solo Chia farming is a lottery system ,
Today's Chia netspace, I figure 200TB have the same chance as 400TB ( 0.00001 vs 0.00002 chance )
Hook up a cable between 2 machines.
configure static IP address in /etc/netplan on both Ubuntu machine
I wish we
Thanks a lot for the great suggestion as usual!
Yes but the lottery is taken every 10 seconds. Emotionally it is a downer but statistically you will win at the end. As you pointed out, you are doubling your chance hence doubling your earnings. Math doesn't lie feelings do
Thanks a lot for the great suggestion as usual!
I now have a similar set-up, the problem is lack my linux knowledge. Plotters don’t have any secondary storage, so I want to mount the plotters to a permanent share on the harvester where the completed plots will be saved. Is there any linux gui tool that can help you set up permanent shares and mount them on the plotter?
I'm sure there is, though I don't know them particularly well myself. However, it's dead simple to do from the command line, so that's what I'd recommend. The Ubuntu Server guide has an easy-to-follow setup guide: Service - NFS | Ubuntu
enable ssh keyless login between two machine
example:
chiaharv1 is the harvester with storage
chiaplot2 is the plotter
From chiaharv1:
ssh-kegen
ssh-copy-id chiaplot2
test ssh without password prompt
ssh chiaplot2
put these lines in a shell script , fetch-plots.sh in chiaharv1
#!/bin/bash
dest=plothd1
chiahost=chiaplot2
rsync -av --remove-source-files \
$chiahost:/mnt/stage1/chiaplot/plot-*.plot /mnt/$dest/chiaplot 2>/dev/null
change your source and dest directory
run this script from crontab
every hour
0 * * * * /home/fun/bin/fetch-plots.sh
every 30 min
0,30 * * * * /home/fun/bin/fetch-plots.sh
to make sure one copy of fetch-post is running
add
pidof -o %PPID -x $0 >/dev/null && echo "ERROR: Script $0 already running" && exit 1
pidof -o %PPID -x fetch-plots.sh >/dev/null && echo "ERROR: fetch-plots.sh already running" && exit 1
pidof -o %PPID -x rsync >/dev/null && echo "ERROR: rsync already running" && exit 1
example:
chiaharv1 is the harvester with storage
chiaplot2 is the plotter
From chiaharv1:
ssh-kegen
ssh-copy-id chiaplot2
test ssh without password prompt
ssh chiaplot2
put these lines in a shell script , fetch-plots.sh in chiaharv1
#!/bin/bash
dest=plothd1
chiahost=chiaplot2
rsync -av --remove-source-files \
$chiahost:/mnt/stage1/chiaplot/plot-*.plot /mnt/$dest/chiaplot 2>/dev/null
change your source and dest directory
run this script from crontab
every hour
0 * * * * /home/fun/bin/fetch-plots.sh
every 30 min
0,30 * * * * /home/fun/bin/fetch-plots.sh
to make sure one copy of fetch-post is running
add
pidof -o %PPID -x $0 >/dev/null && echo "ERROR: Script $0 already running" && exit 1
pidof -o %PPID -x fetch-plots.sh >/dev/null && echo "ERROR: fetch-plots.sh already running" && exit 1
pidof -o %PPID -x rsync >/dev/null && echo "ERROR: rsync already running" && exit 1
- how many of these drives are you using for parallel plotting on 1 machine in raid0??
- and how fast is your write speed (assuming you use them in raid0) ?
I actually never tested them for speed. 8 of them were sufficient enough to make 6850K to hit 100% CPU. NVME path is too expensive to follow so this is not for speed but cost control.
because i am considering to convert this case to as external jbod with 2.5 inch 900gb 10k disks. connect them to LSI raid card with sff-8088 in raid0 mode
CS381
To fulfill the needs of enthusiasts looking for a compact, flexible NAS chassis, SilverStone created the CS381 with unprecedented level of space efficiency. Designed to house a Micro-ATX motherboard, it gives users a wide array of hardware options that includes robust backplanes to support...
www.silverstonetek.com