Hi everyone. Quick question so I don’t miss out on potential winnings… I am using an RPi4 (no timeout issues here) as my full node and moving physical disks over from my plotter, which is also a remote harvester, until disks are full and ready to be moved to the full node. Both are running Linux in cli mode (no gui). I have connected the harvester to the full node, as described in the chia wiki and it “seems” to be working… ie, debug log shows that the haverster is communicating with its peer and also finding eligible plots. My problem is that ‘chia farm summary’ on both sides only shows the local plots and not the remote (or total) ones. This might just be an implementation issue, but I just want to be sure. Up until a few days ago, I had the full node harvesting the remote drives over NFS for a while, and won a block, but the risk of a potential timeout was too big for me, especially while copying finished plots to the disks when they are scanned for eligibility at the same time. As remote harvesting is - at least in theory - the better way to go, I went with that. Does anyone have any experience/info on this? I found two unresolved github issues which seem to be related, but the team is waiting for PRs from someone to implement the gathering of (hopefully only statistical) info from the remote harvesters to the full node. Thanks and sorry for the long post.
Github issues:
For the GUI: Feature request: Update farm overview to include distributed harvester information. · Issue #2830 · Chia-Network/chia-blockchain · GitHub
For the CLI: [ENHANCEMENT] `chia farm summary` should show remote harvesters · Issue #1461 · Chia-Network/chia-blockchain · GitHub
I think, just to be on the safe side, I’ll set up the plotter as a full node, as well…
i have almost the same as you
A rpi4 (Ubuntu) to farm with no HDD’s at the moment. and a windows 10 PC with 2x16TB to save the plots (inside shared folders), that when they full i will put it directly in rpi4.
i configurated the folders shared between the 2 systems, and in rpi configurated chia to add the 2x16TB over the network.
Also i have GUI configurated at Windows PC to check farmer state.
At the moment i have no issues, time are pretty normal
Hello vavi,
thanks for your thorough reply. It’s definitely another piece on information to have.
One thing that I noticed right from the start, was that the Rpi was struggling a bit (higher cpu) with fuse (which is used to r/w NTFS or other non in-(Linux)kernel filesystems).
So actually, while I started plotting on a Windows machine, I setup a second disk on it with ubuntu and plottet with that. I have also been seeing a 15-20% performance uplift in plotting on Linux.
Since then the RPi was running much smoother, even when the plots where just shared to it over the network. However, when plots are being written to the disks, there might be bigger dropouts… that with remote harvesters (especially wih a non linuxy filesystem)
You have to take into account that the 30s include the whole communications turnaround, not only the drive response.
Thanks again, and please let me know any other thoughts or findings on this.
Hi, none that I know of, but there is a pull request on github which seems to be fixing the “counting” issue.
We should be seeing it in a next version, after the PR is is accepted.
yet. Yes, I have been keeping an eye on the logs from the start and it seems to be just fine the way it is. Thanks