Temporarily Pausing MadMax plotter

Hi,
I was wondering is there a way to pause plotting process for like 1-2-3 hours and then just resume it?

Usecase:
Plotter is running 24/7 on rotational HDD, virtually killing its ability to copy file to next disk when I arrive to collect the plots to mobile USB drive.

Please, just don’t tell me “wait to copy option”, “Why don’t you stop the plotter”, “why would you…” etc.

I want to pause plotting process for 3600 seconds and then resume the plotting to have functioning HDD for 1 hour window just to do some coping because while plotting is running my read/write times are 5MB/s and it just takes hours to copy some file while MadMax is MadMax-ing my drive.

I feel for you but I am not aware of any way to pause this process.

If you cannot change your architecture to enable faster copying then I cannot see a way apart from the things you’ve asked us not to suggest above. Have you considered installing a few more drives?

I have a machine which cannot plot directly to the network share where my farm is, so I have two plots folders. I set MM to fill one with 2TB of plots (18), then stop. Then I start these copying, then restart MM to fill another plots folder with 18 plots whilst the first 18 copy, then manually copy and switch. I have 2 JBODs one with 24 x 2TB drives and 1 with 24 x 1TB drives in them. It’s a bit laborious but I don’t want everything fully automated, where’s the fun in that?

If you’re on linux, you can do:
while true; do chia_plot -f xxx -c xxx -t /mnt/tmp -n 1; sleep 3600; done
This will do one plot and a sleep of 3600 seconds in a endless loop.
Something similar should be possible on windows too.

Why is this no option for you? It does exactly what you need!

Easy.
Don’t use -w, instead do this:
@echo off
:loop
*** the usual mad max parameters ***
timeout /t 3600
goto loop

This will stop everything completely for 3600, and then start again, and it repeats.

The thing is, I have like 5 machines that are idling so I put them to work to make plots (they have just 1 drive so nowhere to copy).
They are producing plots 20-30h/plot and I’m returning on random hours every 48-96 hours just to collect finished plots on USB.
If I put to wait 3600s, they will just wait 1h and continue without me being there to collet at that precise hour.
So when I show up to collect at some random hour, I don’t want to stop plotting and lose potentially 15-25h of process if I catch them that far apart. I just want to pause, copy and let them continue their stuff.
I also want to avoid make 1 plot and then wait for me, that can be few days :slight_smile:

It is silly optimizing request but I was wondering if there is such solution. Perhaps others are wondering the same.

@DjDemonD Machines can’t be touched to install new drive (I’m not the system admin). They are just several machines doing nothing so I wanted to use them to do anything.

But thx guys for suggestions. I think best option is to just break the process and create new one after copy. Not optimal but good enough.

I like madmax, but its a resource hog

First of all get it under control

nice -n 5 madmax/chia-plot -r -n -t -2 -d

What nice does is puts it in a lower cpu priority so the cpu cycles don’t hog the system
Then "ps -ax | grep chia| get the PID of the madmax process and “Sudo ionice -p PID -c 3”, this tells the system to only let madmax hog the IO at idle condition. ( -c 3, means IDLE see man-page for ionice )

What I find with my 32core AMD super-computer is -r 4 is ok but -r 8 is death, and -r 16 madmax owns my server ( I use for other things )

One way to pause or slow down is get the PID of madmax and “Sudo nice -p PID -n N” where N > 5, and reduce the priority real low. [ note with nice small number are high priority big numbers are low ]

So in effect this can ‘pause’ because when max is running you can reduce the cpu priority low, where you can essentially cause it to crawl; When u want to return to default just set it back to ‘0’ or ‘5’ as shown, default on linux is ‘20’ the ‘5’ makes it ‘25’.

WRT to effects If I let max OWN my system, then I plot in 20 minutes, with the nice5 & idle io, I do about 35 minutes, but the system is usable.

2 Likes

Excellent @somemoar !
That did not crossed my mind.
This is great solution and it is working on Windows as well.

Can maybe windows or linux make HDD I/O priority because this seems to be a bottleneck.
CPU reduction is indirect help, but direct way would be to nerf the priority for MadMax HDD, if that is even a thing :slight_smile: