I’m going to wait until a few of my plots finish up and then monitor the performance. Then I’ll give that a try.
Performance is improved quite a bit with the cronjob and while NOT plotting. My response times are around 40-90 seconds now. I also noticed the warning message says this and that the time should be below 5 seconds.
2021-07-11T13:55:46.932 harvester chia.harvester.harvester: WARNING Looking up qualities on /Volumes/ChiaP-8TB1/plot-k32-2021-07-09-13-24-95f8bf5ced70beb4b223337622fa1ce82781a4c77128d32722054730fa6a6431.plot took: 45.26976704597473. This should be below 5 seconds to minimize risk of losing rewards.
2021-07-11T13:55:46.933 harvester chia.harvester.harvester: INFO 2 plots were eligible for farming 4abd7d0d3b... Found 1 proofs. Time: 45.27037 s. Total 372 plots
I got about 7 valid ones in a row, followed by 2 or 3 invalid ones, so the success rate is much better.
I also noticed some chatter about needing a newer python version on Mac, but I have 3.9.
Thats quite a bit of improvement ! I still got only 3-4 valid plots today so my success rate is much lower , but still better than without the wake job !
Are you running the 1.2 dmg or built it ? I just now completed a development build ( for 1.2 ) and waiting to see the results … i also tried a build of the main branch ( so 1.2.1 ) but it fails with malformed sqlite db error …
did you also switch hdd s ? am yet to try that
Just running the released version from dmg and a cronjob that touches a file every minute. Didn’t change drives but have plots on 2 different ones.
The next thing to try will be setting up my raspberry pi 4 as a farmer.
Started plotting a single plot again and it killed the response time. This is silly because my computer isn’t underpowered.
i built 1.2.0 on python 3.9 with no improvement whatsoever … also upgraded to Big Sur … the only thing am left to try is to run the farmer in a docker … this is beyond annoying atm
Its resolved now … need to format the external drive to applefs instead of exfat … the partials response is under a second now and all submissions are valid … it would be hassle i think later on if we want to connect the drive to a rpi harvester , not sure how the response would be with apple fs fuse on linux but for a mac harvester this is the way for now …
1 Like
Wow, that’s a huge difference. I am going to try updating to the latest to see if it’s any better, and also plug the drive into my raspberry pi out of curiosity. Just set up a harvester node today. I have a handful of drives to do, so maybe the next one will be reformatted to apple fs.
Ok, so the Raspberry Pi 4 is doing very nicely with the exFat formatted drive and reading pretty quickly:
2021-07-12T21:59:34.445 harvester chia.harvester.harvester: INFO 1 plots were eligible for farming 7ecde9bfb8... Found 3 proofs. Time: 1.64304 s. Total 108 plots
2021-07-12T21:59:42.994 harvester chia.harvester.harvester: INFO 2 plots were eligible for farming 7ecde9bfb8... Found 0 proofs. Time: 0.13126 s. Total 108 plots
2021-07-12T21:59:50.958 harvester chia.harvester.harvester: INFO 0 plots were eligible for farming 7ecde9bfb8... Found 0 proofs. Time: 0.02075 s. Total 108 plots
2021-07-12T21:59:59.492 harvester chia.harvester.harvester: INFO 0 plots were eligible for farming 7ecde9bfb8... Found 0 proofs. Time: 0.05168 s. Total 108 plots
2021-07-12T22:00:09.563 harvester chia.harvester.harvester: INFO 1 plots were eligible for farming 7ecde9bfb8... Found 1 proofs. Time: 1.46609 s. Total 108 plots
I’ll plot for now with drives formatted in the apple format so I don’t miss rewards and then I’ll see how/if the RPi4 can read them.
I’m wondering if SMB would work fine too over the network, plotting and then setting the final destination to a drive hooked up to the RPi4.
I also had the issue of slow response time. I farm with a MacBook Pro.
After reformatting the drive which contained the plots from ExFAT into APFS the issue was solved! Thanks for the tip. Much appreciated.
3 Likes
Usually lookup times need to be under 5 seconds to avoid stales, under 2 would be best. Also the reference pool protocol in Python is quite slow at processing received proofs especially when its flooded with them. Likely the 1st is the reason behind most of your stales though.
I reformatted a drive and now I’m in the under 2 second range. ExFat works terribly for retrieving proofs on the Mac and it should not be used. No issue on Linux though.
2 Likes
Ex fat is not the recommended / fastest on Linux, I can’t tell you what is as I’m on Windows, but I’ve seen it mentioned countless times on the forum in one of the madmax threads.
1 Like