I have something in my head that tells me that ZOS reserves 100GB. Although it sounds a lot ( to me) but I have this number in my head⌠If true your 64GB wouldnât be enough even for the OS⌠Also for 4 cores for maximum earning (gold ratio) you would need 400GB ssd (a little more because you never get 100% out of it)⌠and 32GB RAMâŚ
Dell WYSE 5070 Thin Client 3Node [RESOLVED]
Nice catch guys.
Indeed zero-os needs 100GB for cache memory.
I think some farmers could make it run on less than 500GB of ram, say 400GB of ram, but 64Gb should be too little indeed. Thanks @ParkerS for spotting the obvious!
@elproducto can you try with a 500GB SSD minimum? This should do the trick.
As for @nooba pointed out, there are optimal ratios for farming. Here they are:
What is the optimal ratio of virtual cores (vcores or threads), SSD storage and RAM memory? What is the best optimization scenario for a 3node, in terms of Threefold tokens (TFT) farming rewards?
In short, for peak optimization, aim for 100 GB SSD of storage and 8GB RAM of memory per virtual core (vcore or thread).
For example, a 32 threads (32 vcores) 3nodes would need 3.2 TB SSD and 256GB RAM to be optimal, reward-wise. That is: 32 * 100 = 3200 GB SSD = 3.2TB SSD, and 32 * 8 = 256 GB RAM total.
Adding more GB of RAM would not increase your TFT rewards. You would need more vcores if you want to expand.
NB: This is purely based on reward considerations. Some users might need different ratios for different specific uses of the Grid.
Itâs not âmandatoryâ as it can be played around a little, but too far from 500GB wonât boot I would guess.
Drives that are over 100gb but under 500gb Iâve have mixed results with, I do remember seeing a node on explorer with 240.
Normally I would expect it to say no cache dish found not just hang on registration in progress. Though this may be because the drive also needed a rewipe.
Itâs difficult to say if upgrading the ssd would resolve your issue for sure but I suspect it would.
Hey @nooba I think you got it right. Itâs 100GB of SSD for Zero-OS.
Not all 3nodes have more than 100GB of RAM. Itâs a big number actually.
Thanks all for your assistance. Reviewed boot output. Noticed a few thing listed below. I will continue my testing once I get a bigger drive.
- no debug disk found
- BTRFS: device label devid 1 transid 11 /dev/sda scanned by udevd (266). /dev/sda is the drive in question
- scsi host1: usb-storage 1-1.3:1.0. /dev/sda seems also to be on scsi bus? Could /dev/sda drive on scsi cause issue?
- several wget: bad address âgoogle.comâ. However further down something does get downloaded from google to /dev/null
If I remember well, the google thing not getting it right, right away, isnât a problem.
Itâs just the 3node pining google.com.
This shouldnât be a concern.
Happy troubleshooting. Youâll see after this adventure you will boot a 3node in no time.
Thatâs the beauty of learning curves
Yes, itâs normal to see a couple failures connecting to Google before a success, which is there.
I have one node running with a 128gb SSD, but Iâve never tried less than 100gb.
Once the node is booted and says âFetching realtime node infoâ, you can connect a keyboard and hit alt-f1,f2 to look for any errors.
The btrfs filesystem suggests that Zos is access the disk okay, since thatâs what it uses for its cache partition.
@scott thanks for the additional troubleshooting step.
I did a netboot on a system with 128 GB of storage. I then did the Alt+F2 once fully booted. I saw what seem to be network related issues (errors listed below). Is there any adjustment needed to get past these errors?
First error is constantly repeating. Second error occurs after many occurrences of the the first error. Second error occur after Yggdrasil network setup dialog. I followed the steps to create the farm, so not sure what is up with the second error, I suspect it is related to the first error
error=âcontext deadline exceededâ module=network
error=failed to initialize rmb api failed to initialized admin mw: failed to get farm: farm not found: object not found
Can you try to boot with a USB and the bootstrap image?
If you can get it to work as simple as possible, then you can explore PXE.
Just an idea.
Maybe even just follow the link above from start, and just create a farm, bootstrap from USB, etc.
Changing the boot method wonât be relevant here, since the node has made it through bootstrap and is already running previously downloaded Zos code at this point.
Can you show a screenshot or two of what youâre seeing? Weâll need this to escalate to the dev team if necessary.
The other common gotcha with farms is that they only work on one network. Sometimes folks try to boot to testnet with their mainnet farm, for example. Can you confirm that the boot media/link you generated is for the same network that the farm is on?
In earlier testing, I created a USB disk using test release, and ran into the same issues. However, I decided to use prod efi file release for network booting. Using prod release the 3Node successfully booted.
For network booting I followed steps provide by @TheCaptain, linked below to setup my pfSense router.
Good to know you got it working!
For others reading this, the issue was trying a test net bootstrap image for a main net farm!
Iâll add [RESOLVED] to the postâs title.
Let us know if you have other issues or other questions.
Okay makes sense. To use test Image I would need to create a new farm. Correct?
BTW test release is using version 3.5.0, while prod is using 3.5.2. Screenshots of errors below.
Yes exactly.
Main net:
https://dashboard.grid.tf/
Test net:
https://dashboard.test.grid.tf/
Dev net:
https://dashboard.dev.grid.tf/
Glad to hear your node is online
Iâve also seen some context deadline exceeded
errors after the wait for slaac to give ipv6
message when booting nodes lately. I think it just means the node hasnât received an IPv6 address yet. This clears for me after a while on my nodes.
As for which version of Zos is running on testnet vs mainnet, thatâs a good questionâmaybe thereâs a good reason for this, not sure.
Nice @scott
I will then add 2 new Q&A in the Farming troubleshoot section: (1) @elproductoâs failed to get farm and (2) your wait for slaacâŚ
Been a while since we added Q&A in there. The FAQ should be hungry for more knowledge by now.
I can confirm test release is working for me as well. I had to create a new farm in test dashboard. Update network boot setting to use test EFI file set with new farm ID.
Alright @elproducto! Thatâs so cool.
I think itâs fair to say that you now master deploying a 3node on the TF grid!
Next time, itâll be you helping out a farmer to deploy their first 3node!