Pi4 Nodes improvements - 2 nodes

Node Operators
1

Hi fellow SNO’s!

I’m a node operator for a while now, mainly starting to learn more about linux, docker and networking in general. Started with one node, expanded to two nodes on two 8TB disks but I see a stagnation in growth, in both the nodes. Both relative to their size. The first node wasn’t “full” at the time of adding the second. But the growth rate was enough to make the jump, even though it would eat at the other one, being on one IP and all.

Being a regular observer of the forum, I sometimes read about certain setups that I don’t really understand. I’ll refer to “what I’ve read”. So I wanted to discuss three topic:

    1. Hardware setup
    1. Monitoring
    1. Software/Linux setup

  1. Hardware setup
    I’m running a Pi4 4GB, with an external 2-bay through USB3 (externally powered). The OS is running from a SD card. My main concern is power-usage as that’s the most expensive part of my operation. So this is as slim a relatively stable setup can get for me. I’m not looking for any major improvements, but I’ve been reading that an SSD OS disk can make a difference. However my only way of adding this would be through a USB3 > Sata connected SSD, and that doesn’t sound a like super stable OS setup to me.

  2. Monitoring
    For some reason I don’t really seem to grasp how to add monitoring to my node without burning (or being afraid to burn) too much resources. I read about using an API, to consume this outside of the node. But I can’t seem to find much in-depth guidance. I’m tech-savvy enough but I need some hand holding, exploring this new terrain.

I’ve been reading about monitoring how many request you get to fill and how many you loose out on. I guess this would be valuable to know, and critical to measure any improvements I’ll make.

I’m able to run local monitoring servers, if necessary. I’m looking for something visual and more in-depth than the dashboard.

  1. Software/Linux setup
    I’m running Pi4/Ubuntu CLI os, with the orders-folder on the node disks themselves. Everything mostly out-of-the-box based on the guides that cover RPI nodes. QIUC is enabled. What could I improve to have my node handle more data? I’m sure some linux-master here know a few tricks. I also see on my dashboards the nodes are from-time-to-time not 100% online. Somewhere around 98% or something around that value. Which makes me wonder if the performance is too slow to even be fully “audited/online” 100% of the time.

Hope you can inform me a little bit on these topic to make improvements on my node! Thanks is advance!

2

Check your logs for fatal.

Could be the timeout error.

How are that 8tb drives formated and wich type are they? Smr?
also you could check fragmentation

I agree, but moving logs, dbs and orders there reduces disk io noticeable. Even a high quality Flashdrive could do. (e.g. i use Samsung BAR suitable reformated.)

1 Like
3

Don’t worry, i notice an equilibrium on my first node around 8TB too.

4

I believe with rpi you can’t use an external drive for boot, you are stuck with SD card, but maybe I’m mistaken. But the SNOs here are using the SSD drive not for OS, but for logs and databases, that are none essential for storagenode to function and they eat up resources. And maybe cache. So using one it’s a good thing.
For monitoring, I believe they use Prometheus/Graphana, but don’t know what resources they use and from where they get the data: API or logs, and if and where the processed data is stored for those graphs and usage history. I didn’t bother with them.
The stored data is pretty stagnant in the last 6 months. I see a very slow increase in all my 8 machines.

5

Yes, you are mistaken.

6

As mentioned above, I would check your logs and make sure there aren’t any significant issues. Losing races is normal. If running on Docker I would check the container up time to make sure it isn’t rebooting.

Based on individual reports, some node operators experience much lower ingress compared to others. I think latency can be part of it. I have one node behind a VPN out in Singapore testing high latency impact on the node versus a low latency local node on the same ISP. The high latency node receives around half as much data as the local node. Chalk that up to losing more races. But it does show that high network latency can reduce ingress significantly versus low latency. In this case it would be the route between uploader and your node, and not node to satellite. How busy one’s Internet could also be a factor. And just some ISP’s are slower than others.
Not saying any of this is your issue, just some interesting observations.

1 Like