CMASSE Posted November 7, 2021 Share Posted November 7, 2021 Hi there - I'm running the full stack of Machinaris containers for all the coins - is there a way to reduce RAM usage by any chance, if farming only? My MOBO can only go up to 32GB :(. Also just wanted to say great job on these Guy! Quote Link to comment
guy.davis Posted November 7, 2021 Author Share Posted November 7, 2021 39 minutes ago, CMASSE said: Hi there - I'm running the full stack of Machinaris containers for all the coins - is there a way to reduce RAM usage by any chance, if farming only? My MOBO can only go up to 32GB :(. Thanks! Actually, there may be. Would you mind manually testing it? The idea would be to run some of the forks with the wallet off, either some of the time or all the time. My understanding is syncing the wallet is a nice-to-have, but not a 24/7 necessity the way the farmer is. So please try the following, in one Unraid shell get a "docker stats machinaris machinaris-flax machinaris-hddcoin ..." running to watch each containers resource usage. Then in another Unraid shell, get into a particular fork fullnode container like "docker exec -it machinaris-flax bash", then do a "flax stop wallet-only". This should stop only the wallet, but leave farming running. Check with "flax farm summary" and "flax wallet show". It will be interesting to see what stops updating status-wise in the Machinaris WebUI with the wallet off. I'm guessing the Summary page's Balance and Farmed amounts. Also the Wallet page. However, I expect the rest *should* keep updating. If this manual test comes back positively, particularly if wallet-off state results in a noticeable resource savings as shown in "docker stats", then I can look at automating the wallet being off. Maybe rotating each container to have their wallet in sync catch-up for 1-2 hours a day at most? Just thinking out loud here. Please let me know what you find! Quote Link to comment
mdrodge Posted November 7, 2021 Share Posted November 7, 2021 (edited) Something I noticed this morning. (I hit a chia block over night and I'm with a pool) In my alerts It says Cha-ching! Just received 0.25 XCH 2021-11-07 05:15 Cha-ching! Just received 1.75 XCH 2021-11-07 05:15 Luckily I'd already seen my wallet and wasn't fooled this time. Apart from that I'd like to say how I think what Guy has done here is awesome. (And his jokes crack me up) Update: The Daily Summery Alert says Received ☘️: 2.038838652322 XCH as if I had double farmed it. (0.038 is my pool reward) Edited November 7, 2021 by mdrodge 1 Quote Link to comment
greyday Posted November 8, 2021 Share Posted November 8, 2021 (edited) Is there a guide to optimizing plotting? New to chia and been running it for a few days now and have only plotted 18 plots (averaging around 4 hrs per plot). I mean, it's just running in the background and not using much in the way of resources so no big deal, but I'd like to get it as efficient as possible... So far I've played with thread counts and haven't noticed much of a difference (tested out 12, 15, 16, 24, and 32 so far). Only real difference I noticed was my first plot only took two hours, second 3, and then the rest have been just under 4; I assume this is an ssd trim thing? Running on a 3950x w/128gb ram, temp is a 2tb SATA ssd cache, plots are saved on a btrfs jbod cache of 5x5tb drives, using Mad Max. Edited November 8, 2021 by greyday clarity Quote Link to comment
mdrodge Posted November 9, 2021 Share Posted November 9, 2021 (edited) 20 hours ago, greyday said: Is there a guide to optimizing plotting? with a 3950x etc you should have no problems getting it below an hour. Maybe 40 minutes per plot. a SATA SSD is a bottle neck but if you plot First temp to ram (since you have 128gb) that won't be such a big problem (I actually plot on windows so I'm not quite the person to ask but my buddy DigitalSpacePort will probably help you out. https://digitalspaceport.com/ His YouTube guides are quite easy to understand and he will respond to comments and help you dial it in. Tell him KillaBitz sent you. I can do a plot on Windows in 40 minutes on a 5900x with 64gb at 2666mhz and a Gen4 NVME and transfer it over the network while the next one starts so your system should do at least under an hour. (especially since Windows is worse for plotting) Though it'll need to use most of the CPU cycles to do that. Edited November 9, 2021 by mdrodge 1 Quote Link to comment
mdrodge Posted November 9, 2021 Share Posted November 9, 2021 (edited) @guy.davis do you need to do much with Silicoin since the restart? (Not rushing you or anything just wondering if it's just a port swap and a patch or a massive rework?) At this point I'm quite happy with the collection of shit coins but if there is any more to come then that could work to Edited November 9, 2021 by mdrodge Biscuits Quote Link to comment
guy.davis Posted November 9, 2021 Author Share Posted November 9, 2021 (edited) 41 minutes ago, mdrodge said: @guy.davis do you need to do much with Silicoin since the restart? (Not rushing you or anything just wondering if it's just a port swap and a patch or a massive rework?) Currently, the Silicoin dev team has not properly released the source for their latest release. No, a tarball is not acceptable. If they do, I'll review their changes and consider rolling a new image. Until then, my recommendation is to stop the Machinaris-Silicoin container. Lots of discussion on this in the Silicoin channel of our Machinaris Discord, starting minutes before their "announcement" last night. Edited November 9, 2021 by guy.davis 1 1 Quote Link to comment
greyday Posted November 9, 2021 Share Posted November 9, 2021 4 hours ago, mdrodge said: with a 3950x etc you should have no problems getting it below an hour. Maybe 40 minutes per plot. a SATA SSD is a bottle neck but if you plot First temp to ram (since you have 128gb) that won't be such a big problem (I actually plot on windows so I'm not quite the person to ask but my buddy DigitalSpacePort will probably help you out. https://digitalspaceport.com/ His YouTube guides are quite easy to understand and he will respond to comments and help you dial it in. Tell him KillaBitz sent you. I can do a plot on Windows in 40 minutes on a 5900x with 64gb at 2666mhz and a Gen4 NVME and transfer it over the network while the next one starts so your system should do at least under an hour. (especially since Windows is worse for plotting) Though it'll need to use most of the CPU cycles to do that. I came across his videos before, they're good and informative but not very...concise. Excellent for a start to finish initial setup, but a lot of scanning for someone familiar with Unraid. The biggest difference seems to be, as you suggested, ram, but setting up a ramdisk (instrux for which I came across on page 10 of this thread). That is a complete game changer, it knocked phase one down to 25% of previous plots... 1 Quote Link to comment
mdrodge Posted November 10, 2021 Share Posted November 10, 2021 (edited) 1 hour ago, greyday said: The biggest difference seems to be, as you suggested, ram, but setting up a ramdisk (instrux for which I came across on page 10 of this thread). That is a complete game changer, it knocked phase one down to 25% of previous plots... some trial and error playing with bucket sizes can help a bit to Edited November 10, 2021 by mdrodge 1 Quote Link to comment
mdrodge Posted November 10, 2021 Share Posted November 10, 2021 5 hours ago, guy.davis said: Currently, the Silicoin dev team has not properly released the source for their latest release. Seems like we are stronger without them then. Good Job. Quote Link to comment
CMASSE Posted November 11, 2021 Share Posted November 11, 2021 (edited) On 11/7/2021 at 12:15 AM, guy.davis said: Thanks! Actually, there may be. Would you mind manually testing it?... Flax normal RAM usage: 2.67 GB Flax after running "stop wallet-only" RAM usage: 2.13 GB Immediately after running "stop wallet-only" here is the response to following commands: # flax farm summary Farming status: Farming Local Harvester 257 plots of size: 25.440 TiB Plot count for all harvesters: 257 Total size of plots: 25.440 TiB Estimated network space: 2.496 EiB Expected time to win: 3 weeks and 1 day For details on farmed rewards and fees you should run 'flax start wallet' and 'flax wallet show' # flax wallet show Connection error. Check if the wallet is running at 6761. You can run the wallet via: flax start wallet The UI on machinaris looks fine for Flax, farming is listed as "Active", wallet is no longer synced, and Flax challenges are still ongoing 1.5 hours after running the command: RAM usage: 2.14 GB Same response from the "flax farm summary command" The UI on machinaris looks fine for Flax, farming is listed as "Active", wallet is no longer synced, and Flax challenges are still ongoing 2 days 4 hours after running command: RAM usage: 2.31 GB Same response from the "flax farm summary command" The UI on machinaris looks fine for Flax, farming is listed as "Active", wallet is no longer synced, and Flax challenges are still ongoing Not sure what I make of it - seems like stopping the wallet gets rid of a good chunk of RAM usage, but grows back again with time. Edited November 11, 2021 by CMASSE 1 Quote Link to comment
mdrodge Posted November 11, 2021 Share Posted November 11, 2021 (edited) On 5/21/2021 at 6:25 PM, Pducharme said: @guy.davis Hi! Just couple of newbie questions. For the Plotting SSD, would you use SATA SSD, NVMe on a PCI-e card or NVMe on Mobo ? I guess SATA SSD cheapest, but slower of the lot? For the final Farming, would you all put that outside of the array using Unassigned devices or for the farming part, using the Free space of the array is good enough? They are questions best answered by your wallet and your opinions. NVMe is better. Outside of the array can be good but it depends on what your array is. I built a dedicated rig so my array is only chia and it has no parity but if your trying to do this on part of a system that does other things to (nothing wrong with that) then unassigned devices is grate for that. I think you know the answers to your own questions. I would point out there is a 3rd array option and that's making a 2nd pool. This can be good if you need different options to the main array. It's better than unassigned because you won't need to add each drive to the containers you can just link the pool Edited November 11, 2021 by mdrodge 1 Quote Link to comment
Ancan Posted November 12, 2021 Share Posted November 12, 2021 On 10/20/2021 at 4:32 PM, guy.davis said: Hi, please ensure that you have copied the mnemonic.txt over and restarted the Flax container: cp /mnt/user/appdata/machinaris/mnemonic.txt /mnt/user/appdata/machinaris-flax/mnemonic.txt Then take a look at /mnt/user/appdata/flax/mainnet/log files: debug.log and init.log. Hope this helps! Late reply, but I still can't get flax to work. I've copied the mnemonic.txt to the correct placement, and also deleted the database file for good measure, but there's still zero plots. Running "flax plots check" from the console finds them all though, and "flax keys show --show-mnemonic-seed" shows the correct mnemonic. In the GUI under "workers" the proper plots directories are shown with correct disk space usage reported, but still the summary shows "0 Total Plots", and lots of alerts about "Your harvester appears to be offline". init.log just shows this: Flax directory /root/.chia/flax/mainnet /root/.chia/flax/mainnet already exists, no migration action taken debug.log is attached. debug.log Quote Link to comment
guy.davis Posted November 12, 2021 Author Share Posted November 12, 2021 2 hours ago, Ancan said: Late reply, but I still can't get flax to work. I've copied the mnemonic.txt to the correct placement, and also deleted the database file for good measure, but there's still zero plots. Running "flax plots check" from the console finds them all though, and "flax keys show --show-mnemonic-seed" shows the correct mnemonic. In the GUI under "workers" the proper plots directories are shown with correct disk space usage reported, but still the summary shows "0 Total Plots", and lots of alerts about "Your harvester appears to be offline". Hi, sorry to hear you're having issues with Machinaris. We can probably solve this faster in the Discord, however, what does `flax farm summary` output? How about your Blockchains status in WebUI for Flax? Are you still syncing? Another way to check this is `flax show -s` Hope this helps. Quote Link to comment
Ancan Posted November 12, 2021 Share Posted November 12, 2021 7 minutes ago, guy.davis said: Hi, sorry to hear you're having issues with Machinaris. We can probably solve this faster in the Discord, however, what does `flax farm summary` output? How about your Blockchains status in WebUI for Flax? Are you still syncing? Another way to check this is `flax show -s` Hope this helps. Appreciate your support. I'll post in the flax "farming support" Quote Link to comment
guy.davis Posted November 12, 2021 Author Share Posted November 12, 2021 (edited) Good day! Machinaris v0.6.4 is now available. Changes include: - Farmr - include `farmr` client for optional monitoring at farmr.net - Stor - cross-farming support for this blockchain fork. - Cactus - cross-farming support for this blockchain fork. Thanks to all who helped test and provided feedback! Known-Issue Discovered: Please use ":test" images for Machinaris harvesters of Cactus and Stor to avoid issue on harvester connection to fullnode. Edited November 12, 2021 by guy.davis known-issue 2 Quote Link to comment
AsTuRkInG Posted November 12, 2021 Share Posted November 12, 2021 No news about silicoin? Quote Link to comment
DoeBoye Posted November 12, 2021 Share Posted November 12, 2021 1 hour ago, guy.davis said: Good day! Machinaris v0.6.4 is now available. Changes include: - Farmr - include `farmr` client for optional monitoring at farmr.net - Stor - cross-farming support for this blockchain fork. - Cactus - cross-farming support for this blockchain fork. Thanks to all who helped test and provided feedback! Thanks for this! One issue. Cactus and Silicoin seem to be using the same port. Since Silicoin is not working, I can just disable it for now, but if it ever gets up and running, there will be a conflict... Can we manually assign a new port, or is it something you need to do? Quote Link to comment
guy.davis Posted November 12, 2021 Author Share Posted November 12, 2021 3 hours ago, DoeBoye said: Thanks for this! One issue. Cactus and Silicoin seem to be using the same port. Since Silicoin is not working, I can just disable it for now, but if it ever gets up and running, there will be a conflict... Can we manually assign a new port, or is it something you need to do? I would not recommend running Silicoin right now. They've gone closed-source with their last releases. Would recommend you run Cactus instead, then no port conflict. CC: @AsTuRkInG 1 Quote Link to comment
mdrodge Posted November 12, 2021 Share Posted November 12, 2021 4 hours ago, AsTuRkInG said: No news about silicoin? 3 hours ago, DoeBoye said: Thanks for this! One issue. Cactus and Silicoin seem to be using the same port. Since Silicoin is not working, I can just disable it for now, but if it ever gets up and running, there will be a conflict... Can we manually assign a new port, or is it something you need to do? Silicoin went closed source (see a few posts higher) they also picked a terrible port. we should be avoiding it since we can't see the source Quote Link to comment
DoeBoye Posted November 12, 2021 Share Posted November 12, 2021 1 hour ago, guy.davis said: I would not recommend running Silicoin right now. They've gone closed-source with their last releases. Would recommend you run Cactus instead, then no port conflict. CC: @AsTuRkInG Sounds like a plan! I will leave Silicoin disabled until further notice! Quote Link to comment
Einsteinjr Posted November 14, 2021 Share Posted November 14, 2021 On 11/6/2021 at 9:15 PM, guy.davis said: Thanks! Actually, there may be. Would you mind manually testing it? The idea would be to run some of the forks with the wallet off, either some of the time or all the time. My understanding is syncing the wallet is a nice-to-have, but not a 24/7 necessity the way the farmer is. So please try the following, in one Unraid shell get a "docker stats machinaris machinaris-flax machinaris-hddcoin ..." running to watch each containers resource usage. Then in another Unraid shell, get into a particular fork fullnode container like "docker exec -it machinaris-flax bash", then do a "flax stop wallet-only". This should stop only the wallet, but leave farming running. Check with "flax farm summary" and "flax wallet show". It will be interesting to see what stops updating status-wise in the Machinaris WebUI with the wallet off. I'm guessing the Summary page's Balance and Farmed amounts. Also the Wallet page. However, I expect the rest *should* keep updating. If this manual test comes back positively, particularly if wallet-off state results in a noticeable resource savings as shown in "docker stats", then I can look at automating the wallet being off. Maybe rotating each container to have their wallet in sync catch-up for 1-2 hours a day at most? Just thinking out loud here. Please let me know what you find! I haven't done any testing yet, but +1 to figure out ways to reduce the RAM usage. My max is 64GB. Quote Link to comment
guy.davis Posted November 14, 2021 Author Share Posted November 14, 2021 2 hours ago, Einsteinjr said: I haven't done any testing yet, but +1 to figure out ways to reduce the RAM usage. My max is 64GB. No problem. I'm always on the lookout for ways to improve Machinaris and blockchain farming. Fyi, I'm running ten blockchains on a single Unraid system with 48 GB total. This Unraid system is doing a lot more than just Machinaris, including Plex etc. Still have lots of headroom: Quote Link to comment
rukiftw Posted November 14, 2021 Share Posted November 14, 2021 57 minutes ago, guy.davis said: No problem. I'm always on the lookout for ways to improve Machinaris and blockchain farming. Fyi, I'm running ten blockchains on a single Unraid system with 48 GB total. This Unraid system is doing a lot more than just Machinaris, including Plex etc. Still have lots of headroom: cryptodoge is new and not listed on CA. still in testing? Quote Link to comment
tjb_altf4 Posted November 15, 2021 Share Posted November 15, 2021 2 hours ago, guy.davis said: No problem. I'm always on the lookout for ways to improve Machinaris and blockchain farming. Fyi, I'm running ten blockchains on a single Unraid system with 48 GB total. This Unraid system is doing a lot more than just Machinaris, including Plex etc. Still have lots of headroom: I've noticed it is at its worst when catching up to the blockchain, I cap my farmer container resources to 2cpu (each fork gets its own pair) and 4GB ram. That seems to keep everything under control for me. Quote Link to comment
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.