Open NewSunSEO opened 1 year ago
it's -t
and -d
, lower case
Yes, I use lowercase in the actual script. The UPPERCASE is part of my naming convention for my organization... I use UPPERCASE for important things. This is how my script actually looks:
CUURENT - Since the plots aren't moving, I went back to using this the same directory for the -t and -d: ./cuda_plot_k32 -n -1 -C 8 -S 4 -t /mnt/pool1/ -d /mnt/pool1/ -c xxx -f xxx
WITH Intel DC P3700, but the plots aren't moving from the -t to the -d. I tried moving them with a script & that did not work either. So I'm thinking I have some sort of permission issue somewhere. I'm hoping someone else experienced this & can suggest where to work on this.
./cuda_plot_k32 -n -1 -C 8 -S 4 -t /mnt/ssd_backup/ -d /mnt/pool1/ -c xxx -f xxx
Any errors in the terminal? That's very strange..
No, none. It just keeps running & filling up the -t drive. Yeah, both the -t and -d are write-able, but I am not able to move the plots from -t to -d for some reason...
What filesystem is /mnt/pool1/
? maybe it reports zero free space?
/mnt/pool1/ is an NFS on TrueNAS. Maybe this is something with the permissions of the NFS share...
If it's a permission error you should see messages about "failed to copy ... "
I will stop the current plot creation using the same mount points & try this again & post any errors in terminal.
Thank you,
I let it run for a bit. It is on its 6th plot. I'm getting plot times tonight around 12 minutes on a 3060 Ti, with 6 new 20TB Seagate Exos drives in pool1. The plots are not moving to /mnt/pool1/ and I am not getting any errors.
You never noticed this message?
It's an issue with free space detection, your -d
is reporting zero free space ... or it's really full
Let's see df -h
output
I did not catch that message, but it is not accurate. This is the destination pool. It was full when it was just two drives, but then I expanded that pool & added four more new drives:
pool1 was full around 37T. It got to 42T when I was plotting to it for both the -t and -d
I am trying to plot to pool2 that is a new pool with all new drives:
./cuda_plot_k32 -n -1 -C 8 -S 4 -t /mnt/ssd_backup/ -d /mnt/pool2/ -c xxx-f xxx
there is no chia_plot_sink_disable(.txt)
file in /mnt/pool1/
?
I did not create anything like that myself. Does your software create that?
No, it's a manual feature, mostly used to avoid to write to the root filesystem when disks are not mounted.
Oh ok. I have scripts that auto mount the mount points at startup. Should I add something for this - chia_plot_sink_disable(.txt) ?
it's strange that writing to pool1 as -t
works but not for -d
, the free space check is the same in both cases ...
Should I add something for this - chia_plot_sink_disable(.txt) ?
It's not needed for linux, since it will fail due to permissions anyways.
Maybe it does copy, but it's super slow and the first copy takes forever and you didn't notice?
The Intel DC P3700 in -t is a fairly fast SSD. I left the plotter for several hours sitting, not writing plots & then came back to it. The plots did not move from -t to -d at all. I stopped it with around 6 plots on it.
Are there any *.plot.tmp
files on pool1 ? If yes what size are they?
There have been, but whenever I see any, I remove them from there.
so it's trying to copy right? dont remove them while it's copying.
try again and check the size of these files, to see copy progress
I suspect the copy operation is super slow, for some reason
The two mount points went offline. I made the plotter VM 400 TiB of RAM, mounted /mnt/ramdisk/ and then added that to my mount script & restarted the whole server. Something knocked my two pools offline. I'm trying to fix them. I will try creating 5 plots again as soon as I have this fixed.
Thank you for your help
I am new to Linux & I know this is outside supporting your software, but quickly, does this look to you like I lost both pools & likely need to recreate them? I know I'll lose the existing plots if I do that...
I just deleted the pools & recreated them new. I created 3 new plots. I think you're right, they are copying, but the process is going really slowly... If you have any ideas what could be causing this, please let me know....
Thank you for your help,
The plots were moving really slowly, but then the CLI disconnected from being inactive for 5 minutes & the copying process stopped when it disconnected...
I deleted all the previous plots I had & started over. The disks were empty, they were definitely not full....
I have no idea about TrueNAS sorry, try to ask for help on my discord https://discord.gg/BswFhNkMzY
Hello madMAx43v3r, I've made some progress & have around 170TiB farming so far, but still having an issue. I changed the Intel P3700 NVME drive, from being a share in TrueNAS, to being directly pass-through to the plotter VM. I am getting 2.5 minute plot creations now, but they are not copying from -t to -d. Both -t & -d are writable. I can create plots directly to both the -t or -d directories. But at this point, the plots just stay in -t and they do not move to -d. This is what I am using, testing creating 3 plots at a time right now:
./cuda_plot_k32 -n 3 -C 8 -S 4 -t /mnt/plots/ -d /mnt/pool1/ -c xxx -d xxx
I can't find anything else so far to help with this.
Thank you,
ls -l /mnt/plots total 224319804 -rw-rw-r-- 1 chia chia 76594369368 May 31 08:28 plot-k32-c8-2023-05-31-08-22-f7c6751af4f53a20f4400ea5e998b5eaf14203e95ba1a0222b9ef4c67d430606.plot -rw-rw-r-- 1 chia chia 76585876024 May 31 08:31 plot-k32-c8-2023-05-31-08-28-48870c7d3c49e4396b5101a7c899c440aba8cd8fb1b4143427a1cf419d59dbf9.plot -rw-rw-r-- 1 chia chia 76523226576 May 31 08:34 plot-k32-c8-2023-05-31-08-30-0a67fc25946775c815196b345c274f16098f0224a70b0030a60ec95ba4596194.plot
-rw-rw-r-- 1 chia chia 76594369368 May 31 08:28 plot-k32-c8-2023-05-31-08-22-f7c6751af4f53a20f4400ea5e998b5eaf14203e95ba1a0222b9ef4c67d430606.plot -rw-rw-r-- 1 chia chia 76585876024 May 31 08:31 plot-k32-c8-2023-05-31-08-28-48870c7d3c49e4396b5101a7c899c440aba8cd8fb1b4143427a1cf419d59dbf9.plot -rw-rw-r-- 1 chia chia 76523226576 May 31 08:34 plot-k32-c8-2023-05-31-08-30-0a67fc25946775c815196b345c274f16098f0224a70b0030a60ec95ba4596194.plot
That was the -t directory /mnt/plots. The -d directory has the same permissions set.
I formatted the NVME like this: sudo mkfs.ext4 /dev/nvme0n1. Maybe it needs to be a different format?
Hello, I have added an Intel DC P3700 1.6TB Solid State Drive SSD PCI Express 3.0 x4 to my R720XD server. There are two pools with 6 Seagate Exos X20 ST20000NM007D 20TB 7.2K RPM SATA 6Gb/s in each pool. Before adding the P3700, I was plotting to the same -D directory as the -T. I have run sudo chmod 0770 on the two mounts for the -D & the -T. The plots are being created in the -T but they are not moving to the final -D. I"m really new to Proxmox, TrueNAS & Chia Farming. If anyone has any insight about where to look to fix this issue, I would really appreciate it. I'm guessing there might be a permission issue somewhere, but I'm not sure where to look yet.
Thank you