My 3080 in Windows takes 300 seconds, looks like Linux should do it in about 160 seconds from what I’ve read.
Just need to find the time, but alas there is a lot of other things which demand my time and are more important.
IIRC the current BB plotter crashes if the temp drive fills up, so my application will actually pause the plotter, I can’t do that in Linux as my program will be running on Windows.
That smb:// thing? I had that problem too. Making a system possword in Windows solved the problem.
I use an old intel server with a rtx 1070. Host is ESXi and the guest is ubuntu. I can create one plot within 6 minutes and Rsync needs 8minutes to the final drive.
I think its something to do with that, as I have Kodi on an Nvidia Shield, and that can no longer access stuff on the server.
What do you mean by a system password?
My plot moves were taking about 8 minutes, but I was moving multiple plots at once to different drives.
I had trouble let linux to access Window system. Plotting in under Linux and transfer to windows. Connection would drop and can’t connect again. Setting a windows os password solved it.
Nice script.
post must be 20 letters.
there is also a link to the original file above!
Hello sir. Want to use The Plow for linux. This video is too confusing for me. Anywhere can I find a written instruction? Thanks!
Sorry, I’ve no idea, and that’s the way it goes (you’re somehow just supposed to know) with Linux, which is why I barely use it.
I wrote my own application in Windows to delete and move plots, now I have Linux running on my plotter I’ll run my plot mover from my Windows farmer when I need to add more plots.
Madmax also wrote something which moves plots, but there is no documentation for it either, not that I could find anyway.
Take a look at this one - chia-gigahorse/plot-sink at master · madMAx43v3r/chia-gigahorse · GitHub (both Win and Linux).
I used it one/twice long time ago to do a cleanup (move already produced plots). You could potentially try to put it in a loop to xfr files either from gigahorse or chia plotter (it doesn’t care about the plotter, it only moves files that it sees at startup (my understanding) thus needs to be in a loop to make continuous xfrs).
EDIT: Sorry, it used to copy files to dst, looks like right now it passes them to plot_sink. So, it will require plot-sink to run to do the xfrs as well (plot-sink will not need to be in the loop, just started before plot-copy.
EDIT: also, I got Defender warning about potential trojan when downloading, so that needs to be bypassed. Max doesn’t bother to contact MS to try to have his Win binaries be put on whitelist (there was a discussion about it on his github issues page).
I’ll stick to what I’ve got, but for others they may well be able to make use of it, but they just seem too complicated.
Mine is still a work in progress, has some minor bugs but worked very well.
Well, it’s really not that difficult.
1: download file. Best in your cuda root directory.
wget https://github.com/lmacken/plow/blob/main/plow.py
then edit this file:
nano ploy.py (or vi ploy.py)
In line 24 it starts with the specification of your NVME-SSD
SOURCES = []
Example from my computer
SOURCES = [‘/mnt/NVME’]
Then it continues with the possible targets.
DESTS = []
Example from my computer
DESTS = [‘/mnt/usb-hdd’,‘/mnt/usb-hdd2’,‘/mnt/usb-hdd3’,‘/mnt/usb-hdd4’,]
THEORETICALLY that’s it.
But with me it works a little bit different, because the plotter is offline and only responsible for plotting. So I take my Windows NTFS disks from the Farmer and attach them to the plotter. Plotter then full and then the disks go back to the Windows Farmer.
If you have the same setup, then you still need to go under:
RSYNC_FLAGS delete the two entries “–preallocate”.
Target:
if BWLIMIT:
RSYNC_FLAGS = f"–remove-source-files --whole-file --bwlimit={BWLIMIT}"
else:
RSYNC_FLAGS = “–remove-source-files --whole-file”
Then plow can also write to FAT/NTFS.
Actually everything is super commented out. The documentation is in the file itself.
The whole thing is a python3 script. You start it with:
sudo python3 ploy.py
And THEN it becomes unfortunately very confusing.
lmacken has really put a lot of effort into the functionality.
User-friendliness was probably at the back of the list or doesn’t really exist.
Endless long lines that break at the back even at 1920px and thus become unreadable. The plot name could have been shortened.
e.g. plot-k32-c05-2023-07-2-09-55 abcde[…]fgh.plot
One juggles usually only with 2-4 files/hard disks.
Speaking of:
My plots are done in about 6min. I started with 4 HDD’s.
Then 3… and later I found out that 2 HDD’s are enough without filling up the NVME. But it can be different for you. Faster plotting requires more HDD’s.
You don’t need to change the script itself. HDD’s (or mountpoints) which are not available will not be addressed.
@lmacken:
In case you ever read in here:
Wish list:
Thanks
Thank you very much. Me no computer guy at all. Your instuction is very much appreciated.
I do the same. Plotting is too fast and writing to NTFS in linux is very slow. I have 4 NTFS to accept plots (manual copying). 4 hard dive copying out can’t match the plotting speed. 40-50 MB/sec wrote speed and I think the auctual speed is even lower. I use external SAS array. Normal copying speed should be 150 MB/sec easily.
Here is the setup
Dell R720 384GB 1080gpu. 4*12=48TB plot dump. Ubuntu. Distribute to external sas array. Destination drives are formatted NTFS on other windows machine. Write speed to NTFS is terrible.
How did you mount the hard disks in Linux?
Show your command.
Plug in hard drive into the EMC array. It auto detects hard drive. Need mount the partition. I use the GUI program, just like windows computer management.
unfavorable. Because now you probably do not know how has been mounted
sudo ntfs-3g -o big_writes /dev/sda1 /mnt/usb-hdd
With this I reach around 200-210 MBit’s
Thanks!
But what did I do wrong?
sudo ntfs-3g -o big_writes /dev/sdc2 /media/user/0623d-SAS12TB
ntfs-3g-mount: failed to access mountpoint /media/user/0623d-SAS12TB: No such file or directory
0623d-SAS12TB is the volume name formatted under windows
