Raspberry 3/4 Node Owner - Did you do any optimizations?

Hey,

I host at the moment one Raspberry Pi 3 whit one 4TB Seageate HDD.
This one move in the next Days to my Parents House and i bring up a second Node whit a Raspberry Pi 4 and a 4TB WD Red at my home.

I run on the Pi: StorJ Node, Storj Watchtower, Storj Exporter, Prometheus and at the moment Grafana, but will remove Grafana on the Pi 3.

I have a lot of upload failes.
Did you do any optimazions whit your Rapsberry Nodes?

I overclok my Raspberry a little bit, but dont do more until know:

dtoverlay=pi3-disable-wifi
dtoverlay=pi3-disable-bt

arm_freq=1300
core_freq=400
sdram_freq=500
over_voltage=2

Thanks :slight_smile:

Define “a lot”. I’m running a node on a VM in a datacenter with 10G uplink and see, right now, something like 40% upload failures.

Don’t overclock the poor thing unless you definitely know that it’s the CPU that’s the bottleneck. You do that by measuring, both before and after upclocking, with “perf” or similar tools. (Eyeballing the output of “top” is not a measurement.)

Acting on a request in 0.08ms instead of 0.1ms doesn’t help when you need 20ms to get the data from the disk and another second (or ten) to send them through your uplink.

How can i view the time of Uploading?
And i overclock all my Raspberry at the first time, when they start :smiley:
Normal thing i do ^^

My connection is: 400/40

	                                        min	max	avg	current
System - Processes executing in kernel mode	0.47	94.93	2.26	2.07
User - Normal processes executing in user mode	9.60	156.06	24.04	23.33
Nice - Niced processes executing in user mode	0	0	0	0
Idle - Waiting for something to happen	113.79	388.20	362.50	363.13
Iowait - Waiting for I/O to complete	0	31.33	7.81	6.13


                                                                                                                           ||min|max|avg|current|
| --- | --- | --- | --- | --- |
|Inactive - Memory which has been less recently used. It is more eligible to be reclaimed for other purposes|198.24 MiB|627.41 MiB|560.67 MiB|338.76 MiB|
|Active - Memory that has been used more recently and usually not reclaimed unless absolutely necessary|207.15 MiB|511.21 MiB|250.05 MiB|496.02 MiB|


                            |min|max|avg|current|

|sda - Write time         |0 ns|325 ms|93 ms|197 ms|
|sda - Read time          |0 ns|37 ms|2 ms|37 ms|

========== AUDIT =============
Successful: 600
Recoverable failed: 0
Unrecoverable failed: 0
Success Rate Min: 100.000%
Success Rate Max: 100.000%
========== DOWNLOAD ==========
Successful: 3764
Failed: 149
Success Rate: 96.192%
========== UPLOAD ============
Successful: 18477
Rejected: 0
Failed: 158954
Acceptance Rate: 100.000%
Success Rate: 10.414%
========== REPAIR DOWNLOAD ===
Successful: 1
Failed: 0
Success Rate: 100.000%
========== REPAIR UPLOAD =====
Successful: 241
Failed: 2395
Success Rate: 9.143%

Judging from your name :slight_smile: you also have roughly the same issue as I now have, namely about 90% upload fails. I guess this is expected as the vast majority of the current traffic originates from the US.

My connection is 1000/100, RPi4/4Gb with USB3 connected storage HDD and system SSD. I doubt I can do anything meaningful to affect the physics of the light propagation.
Roundtrip to US is closer to 200 ms. If un-acked transmission window doesn’t cover the whole piece then the roundtrip multiplies.

I’m however glad that the January test traffic originated from Germany, it saturated my (previously 20Mbps) uplink for the whole month and proved I have a good hardware for the job.

Nice to know, that im not alone :smiley:
But maybe someone is here, who has more succesfull Uploads.
Because if all Raspberrys run now whit only 10% Upload, i think its not worth to run them any more.

1 Like

I have 84% of Ingress succees and 85% Egress success on all nodes

Did you change anything? Overclocking, change settings or something?
Can you tell us your Setup and some Informations?

I have Pentium and I5 CPU, and 1 Rasperry pi

More Informations from the Raspberry are great.
Because about that is the topic :slight_smile:

I have a rock64 with a USB 3.0 external drive. Similar hardware. The only optimization I did was when mounting the external drive, I added the noatime option to the fstab entry. This means the OS will not modify the “last accessed” attribute when it reads a file, saving a write for each file read. Not sure if it makes a difference, but every little bit helps.

Also having lots os upload fails the last few days on my Pi 3b+. The pi runs on a 1000/200 network (it only gets 100/100 bc no gigabit interface, but seems more than enough). Basically I only run 2 dockers on the pi, StorjNode and StorjExporter and a openvpn server to have access to the network where the pi is, so no high load on the pi.
My other node running on my main server seems to be fine. (500/100 network)
image

Where are your nodes physically located?

Guys it’s not your PIs… I have a Ryzen 2400 with USB3.0 connection etc and still only have 14% upload success. Living in germany too

2 Likes

I’m from Portugal. I have 1 node on my main server at home (which is working fine, print below) and the pi is at work (<2km from my home). The ISP is also the same, but the ips are in a different /24 subnet.

image

it is in the, Estonia.

So, some problem with German ISPs? I live in Lithuania and my server (not Pi) gets about 73% PUT success (over the last 297 hours since last update).

Thats maybe true. Think all here whit the Problem are in Germany…
So, maybe Stroj Devs are interested in these, can make a pool or something.

I’m in Switzerland but the symptoms are the same.

From Portugal, running 2 Pi´s nodes, one in PT and the other in NL.
Both with upload success rates <10% :frowning:
I believe the network just got more efficient, and the Pi´s are not that good, in addition to traffic slowed down a lot.
It started with release 0.30
Have other 2 Synology´s and 2 Microservers running great with general success rate over 80%.
Pi´s are just to slow!

2 Likes

I also run it on raspberry pi, success rates is very low :confused: Also io wait time is high 4.5, any tips on how to lover it?