HOWTO: Chia Farming on SCALE

[ornias]

I would be very pleased to have plotman included in de chia image. It can now be installed through the shell manually but after an edit it has to be reinstalled again. Plotman's plotman.yaml config file should be on a dedicated mounting point /root/.config/plotman, just like chia's root/.chia and /plots mounting points. Hope it can be done

TrueNAS and iX-systems don't build the containers, they only provide the Install-GUI ;-)

 

[Johan01]

TrueNAS and iX-systems don't build the containers, they only provide the Install-GUI ;-)

Well, then I have to bake my own ;-)

 

[SweetAndLow]

Well, then I have to bake my own ;-)

You will probably want separate containers for plotman. This way you can continue plotting when you have to update your farmer/harvester.

 

[leckyBill]

I know, scale is alpha, charts are WIP, and this chia chart was probably just an experiment but am I missing something with this chia chart?
The entire thing runs on a single pod... is any interface exposed outside the kube? Is the only way to interact with the farm by web UI shell?
How does one connect truenas cli to pod? lack of bash history and no copy paste makes it really difficult to use...


I see mention of additional tools like plotman and kind of expected there to be multiple nodes inside this chart. Any success adding UI interface to a chia farm on tn scale? Maybe I'm just deluded or green, but I really thought a chia farm would be the ideal scalable application here...

 

[gwaitsi]

are there any instructions for getting this running in a jail.

 

[ornias]

are there any instructions for getting this running in a jail.

Thats off-topic, because this is the SCALE forum. I suggest making a seperate thread about this accordingly...

I know, scale is alpha, charts are WIP, and this chia chart was probably just an experiment but am I missing something with this chia chart?
The entire thing runs on a single pod... is any interface exposed outside the kube? Is the only way to interact with the farm by web UI shell?
How does one connect truenas cli to pod? lack of bash history and no copy paste makes it really difficult to use...


I see mention of additional tools like plotman and kind of expected there to be multiple nodes inside this chart. Any success adding UI interface to a chia farm on tn scale? Maybe I'm just deluded or green, but I really thought a chia farm would be the ideal scalable application here...

Adding to or Adapting the Apps would require you to make your own. Thats the TLDR about it.

 

[Johan01]

I know, scale is alpha, charts are WIP, and this chia chart was probably just an experiment but am I missing something with this chia chart?
The entire thing runs on a single pod... is any interface exposed outside the kube? Is the only way to interact with the farm by web UI shell?
How does one connect truenas cli to pod? lack of bash history and no copy paste makes it really difficult to use...


I see mention of additional tools like plotman and kind of expected there to be multiple nodes inside this chart. Any success adding UI interface to a chia farm on tn scale? Maybe I'm just deluded or green, but I really thought a chia farm would be the ideal scalable application here...

Currently my way of working is to

1. login to TrueNAS system as a user,
2. start a tmux session ("tmux"),
3. create 6 windows for parallel plotting (Ctrl-B c),
4. in each window open a shell to the chia pod (Ctrl-B 0, "sudo docker exec -it container_id sh", Ctrl-B 1, ....),
5. start staggered plotting in each window, copy and paste works
6. detach from the tmux session and leave it plotting in the background (Ctrl-B d),
7. now and then check the plotting (log as user, attach to the tmux session: "tmux attach-session -t 0").

 

[leckyBill]

Currently my way of working is to

1. login to TrueNAS system as a user,
2. start a tmux session ("tmux"),
3. create 6 windows for parallel plotting (Ctrl-B c),
4. in each window open a shell to the chia pod (Ctrl-B 0, "sudo docker exec -it container_id sh", Ctrl-B 1, ....),
5. start staggered plotting in each window, copy and paste works
6. detach from the tmux session and leave it plotting in the background (Ctrl-B d),
7. now and then check the plotting (log as user, attach to the tmux session: "tmux attach-session -t 0").

Hi Johan01,
That's interesting. I was trying to attach with k3s kubectl, but that was not working (k3s kubectl exec doesn't seem to work either) runs, no error, but also doesn't do anything, just ends...
I've actually been attaching thru the UI to shell and installing tmux in the pod and doing the same, except the plotter is running straight in the tmux window, that way, not losing the pts if anything happens to the sh.
I think I will try running sudo docker exec with the tmux + plot command in one... :D
I know it will be largely system specific, but what stagger time do you use?
I was experimenting with parallel plotting trying to find a balance between power consumption and plot time. recently trying ramdisks, but plots still take 7hrs20 to complete and not parallelizable...
Thanks for the tips with the docker exec. can also check farm summary now without a browser :D

 

[Johan01]

Hi Johan01,
That's interesting. I was trying to attach with k3s kubectl, but that was not working (k3s kubectl exec doesn't seem to work either) runs, no error, but also doesn't do anything, just ends...
I've actually been attaching thru the UI to shell and installing tmux in the pod and doing the same, except the plotter is running straight in the tmux window, that way, not losing the pts if anything happens to the sh.
I think I will try running sudo docker exec with the tmux + plot command in one... :D
I know it will be largely system specific, but what stagger time do you use?
I was experimenting with parallel plotting trying to find a balance between power consumption and plot time. recently trying ramdisks, but plots still take 7hrs20 to complete and not parallelizable...
Thanks for the tips with the docker exec. can also check farm summary now without a browser :D

Hi Bill, hope the tips works for you (and anyone else). So in step 1 I don't use the web UI shell but use ssh to login to the machine or do it locally. Stagger time is about 30 minutes. I use two NVMe M.2 SSDs and four 12Gb/s SAS drives as individual stripes for tmp filesystems which I made available to the pod in de settings as host path volumes. This feature has been made available in the latest version. The NVMes do one plot in around 8 hrs and the SAS drives do 9-10 hrs. Systemload is around 30%. When I have more time I would like to integrate plotman in an dedicated Chia image to be more efficient and for easy management.

 

[linbing]

new chia-plotter mutil thread polt tools

GitHub - madMAx43v3r/chia-plotter

Contribute to madMAx43v3r/chia-plotter development by creating an account on GitHub.

github.com

first: build "chia-polt" in other linux system , then upload to truenas scale.

after: you can ssh to truenas, edit a xxx.sh to run


[P4] Starting to write C1 and C3 tables
[P4] Finished writing C1 and C3 tables
[P4] Writing C2 table
[P4] Finished writing C2 table
Phase 4 took 72.5964 sec, final plot size is 108798234709 bytes
Total plot creation time was 1819.22 sec (30.3204 min)

 

[Johan01]

new chia-plotter mutil thread polt tools

GitHub - madMAx43v3r/chia-plotter

Contribute to madMAx43v3r/chia-plotter development by creating an account on GitHub.

github.com

first: build "chia-polt" in other linux system , then upload to truenas scale.

after: you can ssh to truenas, edit a xxx.sh to run


[P4] Starting to write C1 and C3 tables
[P4] Finished writing C1 and C3 tables
[P4] Writing C2 table
[P4] Finished writing C2 table
Phase 4 took 72.5964 sec, final plot size is 108798234709 bytes
Total plot creation time was 1819.22 sec (30.3204 min)

Thanks, read and heard about madmax before, will try it soon. Have you tested and observed the performance when doing (f.i. 6) plots in parallel?

 

[Etorix]

Madmax plotter always takes ALL of your resources. You do one plot at a time, as fast as possible, and then move to the next one. Best results are achieved if you can make a 110 GB ramdisk for use as temp2.

Some results on Ubuntu:
Xeon Platinum 8255C (24C@2.5 GHz) + ramdisk = 20-21 minutes per plot, 72 plots/day
(with Swar plot manager, optimised chiapos and two SSDs, I managed ca. 57 plots/day, each taking 6 hours; a third SSD might have brought me to 70+)

Xeon D-1541 + ramdisk = 55 minutes/plot!
(versus 21 hours for a single plot with optimised chiapos, not really worth considering)

Xeon E-2278G, no ramdisk = 42 minutes/plot, 33 plots/day
(Swar+chiapos: 26 plots/day on a single Intel P4500 SSD, which is not the fastest in my assortment; also 6 hours per plot)
Without a ramdisk, productivity is not much improved over what could be achieved with the standard plotter, but there is no need for multiple SSDs, maintenance does not require planning 5-6 hours ahead and finding the right parameters is much, much, faster (4 hours of experiments vs. 4 days to set a stable regimen for the 8255C).

 

[leckyBill]

Tried with 2 Ramdisks 220G and 110G with all 56 threads and final destination to ssd. When running on box booted from vanilla ubuntu 20.04 it was plotting (including write to disk) in ~24mins
Running the same on a truenas scale box( standalone plotting, harvesting still done in chart) the same process too 35mins and writing to slower final disk 39 mins in total per plot.
I don't know how to reserve ramdisk so that other processes don't try to use it... Did 5 plots in a row no problem, the new plotter starts the next plot while it is still writing the final of the previous one so overall it's even more efficient, but when I tried to run 10 in a row, got the plotter crashed on the 7th... I suspect the ramdisks became "full" as it was the same error when trying to restart without emptying a bunch of tmp files after the crash.
Truenas dashboard showed 75-80% CPU and about 400G memory free... (which it wasn't)
Top showed ~5000% CPU utilisation which is great but not sustainable in a home lab in summer time

 

[linbing]

That's crazy， GPU chia plotter

GitHub - achow1o1/chia-plotter-CUDA-OpenCl: chia plotter CUDA, OpenCl

 

[FreeZM]

Reddit - Dive into anything

www.reddit.com

And once again - mankind is destroying the good ideas behind blockchain. Or must one say that the regulations are once again absolutely inadequate. Error in design.

 

[Johan01]

Today I got some time to play wih madmax on my truenas box. 40 mins is the fastest I can achieve with 2x Xeon silver 4210 CPUs, 20 threads of 40 in use, a 128GB ramdisk for -2 and for -t a 1TB corsair MP600Pro NVMe. I tried with 32 threads but it didn't speedup the total time.

 

[Kris Moore]

Nice to see lots of other Chia farmers doing it on SCALE as I am :)

I'll be working to update the plugin shortly and including tools such as MadMax, Chiabot/farmr and plotman. If there's other tweaks I can do to the plugin to make it better, please let me know.

 

[Johan01]

Nice to see lots of other Chia farmers doing it on SCALE as I am :)

I'll be working to update the plugin shortly and including tools such as MadMax, Chiabot/farmr and plotman. If there's other tweaks I can do to the plugin to make it better, please let me know.

Hi Kris, thanks for all your efforts! I'm looking forward to the official launch of v1.2.0 in the upcoming week which includes the pool protocol. I'm already testing the standalone GUI on mainnet with xchpool now, this afternoon we got a winner already . The plots are made by the madmax plotter with the latest -c pool contract option. I noticed that plotman is not working on a pool implementation yet. Would be good to have version 1.2.0 for the app.
Cheers!

 

[Jeremy]

TruePool MainNet pooling servers are LIVE for testing. Our website has also been updated with live MainNet space graphs and farmer leaderboards, with more updates planned in the coming weeks to provide even more transparency on pool server operations.

We are all eagerly awaiting upstream Chia to publish version 1.2.0 with official pooling support, however early testers and adopters can find information on how to get started right now on our Join TruePool page. As with any early access, there is a slight risk of last second changes to the plot format before 1.2.0 is officially tagged, so please use your discretion before dedicating all your plotting resources too soon.

Join our Telegram Channel

Join our Discord Server

Follow Us on Twitter

 

[ornias]

TruePool MainNet pooling servers are LIVE for testing. Our website has also been updated with live MainNet space graphs and farmer leaderboards, with more updates planned in the coming weeks to provide even more transparency on pool server operations.

We are all eagerly awaiting upstream Chia to publish version 1.2.0 with official pooling support, however early testers and adopters can find information on how to get started right now on our Join TruePool page. As with any early access, there is a slight risk of last second changes to the plot format before 1.2.0 is officially tagged, so please use your discretion before dedicating all your plotting resources too soon.

Join our Telegram Channel

Join our Discord Server

Follow Us on Twitter

Great idea guys!