Just curious. My Estimated Time to Win is currently 14 days. During the dust storm (and recovery from the dust storm), it dropped to 12 days TTW. I had minimal issues (compared to others) during the storm. Just random missed signage points and such. Never lost sync as far as I saw. But then the day the storm stopped (I stopped seeing the effects at 0500 EST on 11/1) I found two blocks. One at 0944 and another at 1505. Before this, my last block was on 10/24. Of course this could just be luck. Just wondering if this could be that my system is far from baseline specs (it used to be a plotter so it is built pretty good) and was able to stay online more during this attack and increased my odds. Thoughts? Is this ammunition to stay away from low spec farmers (like pi’s) or was I just lucky?
I have also won 2 blocks within 2 days of dust storm with ~4.8k plots.
The system is 6c12t i7 8700 with 128GB RAM, I am slowly plotting on it with using only 5 threads. During normal operation the stale rate is extremely low. During storm got hit like everyone else with whopping 75% stales.
My theory is nodes of peers on the network were unable to keep up and couldn’t win any blocks. So folks running better nodes ate their lunch.
It can happen boths, low specs farmer went out, network size decreased and you have higher chance to win. Congrats! Crying in my rasperry4
I heard in a discord chat one of Chia dev team mentioned they might publish a list of minimum system requirements for running full node.
I’m sure Pi will handle well as a harvester. Syncing full node though requires a performant SSD and good CPU, so better leave it to a desktop/server rig
Yes, bad code is always easy to explain with low performing HW. Just check the other thread, where the guy was fighting his 80+% stale ratio and ~5% farm size reporting trying to find problems with his box/connection. After switching to Flex client, his stales went down to 0.04% (just one stale instance). Nothing else was changed in his setup. Again, this is not shilling for Flex, rather reporting the difference in results with two different clients, where only one is well written.
Also, as far as that attack, there were two different problems. Those boxes that were behind (most likely with blockchain db on slow drives, nothing else; although, one can argue that that db is not indexed right), and those that were asked to send bazillion of bytes to their peers that were behind, and thus choked their upstream bandwidth (thus those stale partials, netspace dropping).
My node was in that second category. Everything else, but upstream bandwidth was running basically with zero resource usage. Chia code didn’t prioritize blockchain vs. updates traffic, neither did any network back-off protocol (what has been admitted by them).
I am not that lucky. My estimated time to win is 15 days. During the dust storm, it dropped to 13 days. I never lost sync, too, and didn’t miss any signage points as I realized. But there is none found block. Also, I haven’t found any blocks for about two and a half months. Look-up time is under 500 ms in all harvesters. I don’t understand what I am doing wrong. Or I am just unlucky :((.
This is not an apples to apples comparison. FlexFarmer is just a harvester that connects to full node run by Flexpool in a data center. Chia software runs full node on your machine.
In this case, I take those apples any day. Sure, that is a bit apple to oranges, but if something smells bad, it just smells bad
If your 500TB farm is recognized as 5-10TB only, and even with that small size has 80% stales (using Chia code), but after a switch to a different harvester you get to be recognized as 500TB with 0.05% stales, how do you classify that difference? Why do we need to waste our times looking for those imaginary hardware problems, where we know what is the culprit.
I am also sick and tired of helping people to recover from corrupted dbs (what was reported before I joined, and zilch was done to fix it), upgrades nonsense, not explained log entries. I really do not understand why Chia is not blocking downloads of bad releases. During this storm, v1.2.10 should be blocked, as that just compounded the problem.
You just want to brush it all under the carpet, so your holly image is still intact? I think we are both software engineers.
Also, the official guy has admitted that so far they were doing nothing to fix issues. As Aspy68 suggested, their only goal right now is their banking business. I tend to agree with that.
Are you running chiadog? I have found that is the best way to find issues. It lets you know right away when something isn’t right. I run it on my full node and each harvester.
Folks running FlexFarmer also reported a dip in effective netspace, comparable to standalone nodes. This is because Flexpool node in data center got hit the same way as everyone else. Maybe to somewhat less extent, assuming their hardware is top tier.
In other words, your harvester may produce 0.05 stales, but full node in the cloud will still have 70-80% capacity issues. Based on what I read in the discord, it is due to weak peers. So you can have the best hardware and internet connection and have no problem syncing, but your peers will provide you most of the stales.
I also had some great luck during this. I found 4 blocks over the last week. And I had only found 2 in the 3 months prior. I went from about 17 days to 15 days estimated to win. Had I known my luck might have improved to that level, I would have moved out of a pool and just farmed myself. I would have more XCH if I had.
Try setting up the farmer on another computer, and only running your tower as a harvester+plotter. The storm hit you hard because your CPU is busy with plotting (my guess).
His point is that running a harvester is not as demanding on hardware as running a full node/farmer, and people who run a full-node and are underspecced got hit the hardest. Flexfarmer more or less behaves like a harvester only. You’re splitting hair at this point, he’s right and his point is valid. Even the Chia team has aknowledged that they need to work on optimization for the full node, quicker than they planned to do it, which confirms this issue.
Of course flexpool can still have issue with their own node, but I’d rather trust flexpool to work 24/7 on their setup than me, a small node, doing (what is supposed to be mostly) passive farming from home.
I was so damn lucky, I went solo 2 days before the dust storm as I was heading out of town and thought I would try my luck while away.
I’m currently at 538TB and had only hit 2 blocks in the last few months since starting. Came back last night to see I hit three blocks within 4 days span. Time to win also went down to 13 days.
People like me must be asking the same question… When’s the next storm!
As you can see in the picture, I created a dashboard with a warning section. It works by showing the warning logs in the log file. I haven’t installed chiadog before, but yesterday I installed both in farmer and harvesters. I’m following, and I didn’t see any errors in 24 hours. Thanks for your advice…
It is always safer to keep any third-party software out of your boxes. For ChiaDog, you can share (read only) just your log file folder, and ChiaDog will be happy with that. Works like a charm for me this way.
Also, in case your farmer or harvesters have serious issues (e.g., crash), ChiaDog running from a different computer would still be alive, and reporting back to you.
Same for me. I found 2 blocks in 3 days with 1500 plots during the dust storm. Stale rate was %1.