search

madMAx43v3r / chia-gigahorse

223 stars 31 forks source link

plot_sink improvement #73

Open aramod opened 1 year ago

aramod commented 1 year ago

When mounting a new disk, you need to turn plot_sink off and on when writing ./plot_sink -- /mnt/d Is it possible to make plot_sink check every time d. I have a lot of hdd connected via jbod, and when I connect everything, the plot takes a very long time, so I have to mount it one by one.

aramod commented 1 year ago

And I would also like to make plot_sink with the removal of old plots <= 100g in turn. It would save a lot of time for plot replotting. HDD space is less than 70gb, but there are *.plot files larger than 100gb - delete this file and write a new gigahorse plot

Such a scenario would make it easier for me to keep track of the farm.

madMAx43v3r commented 1 year ago

I don't want to provide any auto-delete of plots, it's too dangerous

madMAx43v3r commented 1 year ago

if you only check for > 100 GiB you already lost all your new k33+

madMAx43v3r commented 1 year ago

HDD space is less than 70gb, but there are *.plot files larger than 100gb - delete this file and write a new gigahorse plot

This is really bad, HDD performance is quite low when full. You will copy at 10% of the max speed at best.

aramod commented 1 year ago

Yes! I didn't think about it.. Thank you!

aramod commented 1 year ago

Is gpu_plotter doing 512buckets at the moment? Is it possible to change?

madMAx43v3r commented 1 year ago

Just mount all your drives and include them in plot sink command line, then delete all plots on certain HDDs over time.

madMAx43v3r commented 1 year ago

Is gpu_plotter doing 512buckets at the moment?

256 buckets, that's hard coded, cannot change it. It's the best performance anyways, 128 and 512 are both slower.

aramod commented 1 year ago

Just mount all your drives and include them in plot sink command line, then delete all plots on certain HDDs over time.

I do. But I have too many machines on linux and windows and I don't have time to follow)))

madMAx43v3r commented 1 year ago

Just do one machine at a time? Or multiple at a time?

aramod commented 1 year ago

Due to the 2x1g network, I have to do sections on linux and transfer to 2xwindows at the same time. Aggregation lan does not work on windows. I have 12 such machines and 4 linux

aramod commented 1 year ago

Is gpu_plotter doing 512buckets at the moment?

256 buckets, that's hard coded, cannot change it. It's the best performance anyways, 128 and 512 are both slower.

The readme says that 512 buckets for chia, and 256 for MMX. How to count correctly for Chia?

madMAx43v3r commented 1 year ago

I think you are confusing something? You mean plot filter?