Let us know how it goes! The memory won’t be a problem on a new node during the current low traffic, but it can definitely become a problem during higher traffic. Especially if that HDD is indeed SMR and transfers can’t be written to the disk fast enough. I hope we’re wrong about that. But either way I’d love to hear how this experiment goes!
4 Likes
Thanks for replaying to me. And that replaying to me mean that you have a double sentence on what you are saying. It’s me that need helping others or others that need to help me?
I’ll give here my experience so far. I’m now close to 1.6 TB used out of 5 but I allocated 4 TB only from my hard drive. I would say it struggle a little. I got almost the same amount of trash than new data. So having new data are getting harder and harder.
About network performance, OpenWRT firmware will give first priority on packets traversing through. What pass through the kernel are done first. DNS service though where I think is not kernel take longer to respond because it share the same priority with storagenode.
The memory usage is very little without using a docker. I’ll guess for about less than 20 MB. I would say that docker are very hungry in resource in contrary without it.
That’s it for now I guess…
1 Like
Stumbled across this thread and wondering how this experience panned out in the end? 
I’m curious to know too, because it’s quite stunning if you could run Storj even on low-edge devices like routers or at least quite less performant than a Raspberry.
Perhaps it’s still working, because from the usage it doesn’t look like a resource consumer, if you use a one CMR HDD with ext4 per node
CONTAINER ID NAME CPU % MEM USAGE / LIMIT MEM % NET I/O BLOCK I/O PIDS
f41a7362682e storagenode2 7.74% 134.1MiB / 24.81GiB 0.53% 75.4GB / 50.8GB 0B / 0B 53
a73bbdf8a1ed storagenode5 4.40% 121.8MiB / 24.81GiB 0.48% 57.6GB / 13.1GB 0B / 0B 37