What is happening with my plotting?

I’m on Flexpool.io and seeing a recurring pattern of Effective space and reported points dropping and rising. It’s been happening the last 3 days and it all started after I upgraded my machines from 1.2.9 to 1.2.11. Has anyone else seen anything like this?

image

I don’t see that much drop on my side (maybe 10% or so).

If you check chiapool.directory, there are two things that make me nervous. Their luck ratio is rather bad (for the past few days). Also, looks like Flex is losing net-space (below 400PB right now).

@Chris22 do you see those things on your end?

Every pool is losing space right now due to an issue with the network.

I also just noticed that my number of peers dropped to 10 (including my farmer and wallet). Although, none of those is blocked. I am both getting and sending data to those peers. I also don’t see stales, so this behavior looks like a new one.

Is that another form of dust storm? If that is CAT1 related, then based on what Chia said, there are only few key nodes running v1.2.8/9, so that would be just a drop in the bucket.

The issue continues. I’ve tried restarting the software, restarting the computer, downgrading back to 1.2.9 which was working before and a reinstall of 1.2.11. Nothing has made any difference. Both of my nodes are on different ISP’s and the strange cycling up and down continues.

Just a Bump for an unresolved issues. How do I troubleshoot this?

Those problems lasted a day or so on my end, and were not that severe as yours, as such looks like not related.

Just to clarify, your both nodes are in the same location (same local network) and are connecting through two different routers (each pointing to a different ISP)? If that is the case, any reason to have two full nodes?

Both Nodes are in different places. 1 at home with 1 ISP. The other is at my office different city, different ISP.

Sorry, I don’t have a clue how to bite it, then.

From what you have described, the only thing in common is your pool (Flex). So, maybe it is a good time to jump on Flex discord channel, or push Chris here.

One more possibility would be that your nodes are somehow underpowered (thus sensitive to traffic fluctuations), but again, seeing two nodes doing the exact same thing would be really a big coincidence.

By the way, did you try to install farmr? Also, pools are using the number of submitted partials to estimate farm size. Maybe installing a heat map monitor would show something? I am using PSChiaPlotter heat map module, but there is at least one more such tool.

Sorry to be monothematic, but whether those heat maps would be steady or fluctuating in a same way as your pool output, that would indicate that either your nodes cannot properly receive challenges (heat maps would fluctuate) or cannot send them to the pool (no fluctuation). Both cases rather point to your nodes not able to handle that traffic. I highly doubt that this has anything to do with your ISPs or local network, but rather boils down to blockchain db not being handled right, thus causing problems for other communicating processes that face the network (chia stated before that it was the case, and should be addressed, whether it was or not, I don’t know, but would not put my money that it was). If that would be the case, then the standard things could be at play: 1. your dbs are on slow media, 2. you need to drop the number of peers.

There was another thread, that the solution was to drop down to v1.2.9, but you have already tried it.

Of course, as you are already on Flex, another option would be to switch to Flex farmer. Potentially, that could be the easiest solution.

1 of the machines started out as a high end gaming PC. The other as a media server with Ryzen 3200g. So I can’t imagine those would qualify as under powered.

I’m a little nervous about using a 3rd party farmer atm, but I’ll take it under advisement.

I assume you are talking about the ChiaHarvesterWatcher? I started it up and I’m all greens and blues.

The Pool tab on the Chia GUI looks very stable compared to what I see on the pool

Sorry for that. There are two definitions of under-powered. Either it is a shit hardware or shit code. As we know, Chia stance is that most of us run shit hardware.

It really boils down to a bad code around blockchain db. As mentioned above, NVMe and limiting the number of peers can provide some relief to that code, but it looks like we are close to what H/W upgrades can do right now as the blockchain grows every day further making it more difficult for that code.

Before you install it, I would suggest that you switch your setup to use a cold wallet (whatever cold means - basically no wallet keys on that box).

Yes, that is what I am using. I run it with 2 seconds test (-MaxLookUpSeconds 2). Those greens and blues further indicate that your harvester is sound.

Again, I would suggest that you try limiting the number of peers for a day/two, and if that will not help, jumping on Flex discord channel.

What would be the process for limiting peers and what would be a reasonable number of peers?

In "full_node’ section look for that line:
target_peer_count: 80

Some people were syncing from scratch setting it down to 10 peers only without any issues. From what I see looking at my peers (Connection tab on Full Node), my node mostly receives data from just a handful of peers.

So, as that is testing only, I would go with 10 peers, and keep it like that for the duration of your fluctuation (e.g., one/two days). If those fluctuations will go away, we can assume that at least this is one of the contributing factors, so I would try to bump it up to 20 or so, and still monitor it for the next “normal” fluctuation, and eventually leave it like that. This way you will be seen as a bit more friendly peer.

There is an ongoing debate whether the number should be 80, or could be lower, but what that debate is missing (IMO) is thah if you have less peers, you just provide them with more data, so your upstream bandwidth is about as much saturated as with higher number of peers. Chia is not really doing much of bandwidth control, rather tries to saturate your upstream bandwidth, regardless of the number of peers.

Second thing is that syncing takes about 2 Mbps only (downstream for you), that implies that from your point of view, you just need one peer to sync properly (assuming that an average upstream bandwidth is 5Mbps or so). Therefore, 10 is a good number to start with.

Oh, and that value is read only at chia startup. So, you need to fully restart it.