I have 120 16TB hard drives per Harvester, currently in pass-through mode, which means I need to wait for the 120 hard drives to be mount/unmount when I turn on/off(reboot) the harvester. However, I have noticed that every time I turn on/off the harvester, I will not be able to fully mount/unmount these 120 hard drives due to timeout.
I tried to adjust the waiting time to 10-20 minutes, but it was too long. Is there any other way?
1 Like
I hope you can find an answer here.
Will be following out of curiosity as I’ll be adding many drives soon as well
1 Like
are you mounting them with -nofail ?
2 Likes
Don’t have them mounted by default when you start your computers, mount them after startup, then start your harvester script(s). Then when you want to turn off, unmount first then chia stop all -d etc.
You have a big farm you shouldn’t be starting and stopping very often methinks.
2 Likes
yes
LABEL=1-1 /mnt/1-1 auto nosuid,nodev,nofail,x-gvfs-show 0 0
The mount option is generated through the Ubuntu gnome disk utility.
2 Likes
Yes, you’re quite right. I may need to create a startup script.
I am still in the process of deployment and do not have any plots in these disks, fortunate to have discovered it early.
I am also testing the RAID scheme.
1 Like
Please wait for more experienced replies too! I am only running two JBOD’s - 5300 C8 plots. I don’t have any scripts for starting and stopping and this is only what I could think I would do.
1 Like
You can adjust the timeout setting in /etc/fstab. e.g.
UUID=c053ffe8-1c9b-4242-bcfe-680da0dbab02 /mnt/disk1 xfs defaults,nofail,noatime,x-systemd.device-timeout=5min 0 0
That’ll wait 5 minutes for disk1 to mount. Adjust as needed.
1 Like
I am currently using RAID and HBA modes for comparison. Before the official version of the CUDA is released.
GPU is performed at a fast speed of 3070 3 minutes per plot with c0. The plow performance is too low, and rsync over ssh is too slow.
Currently, there are 12 hard drives in one array, with two Harvesters in HBA mode(120 hdd/harvester) and three in RAID mode.
Until the official version of BB CUDA is released, the final plan will be decided.
