If you could do this: wouldn’t it mean a SNO is exerting control over when a paying client deletes their files? Like that client is saying “please store this, I’ll pay you, and delete it in 3 months”… and you say “no, I won’t store it if you want to delete it in 3 months”?
I could see the dashboard (or scripts) maybe getting upgrades to report on TTL info. But SNOs shouldn’t get to pick-and-choose the data they store (if they want to get paid).
(but… if your node magically lost every race to accept TTL data… I don’t see how Storj would know if that was intentional or accidental… you’d just make less money)
choiceofn with n=6 has broken a new record and was about 20% faster than we would need. For some reason not for long. The performance decreased. We will now run with a bit less load for a longer time to see if it is stable.
How does it look on your side? Do you see any issues why the performance might decrease?
I don’t know why my nodes at home aren’t getting that much load. I think my cgnat and the VPN from portmap is the problem. Does anyone know a better way to get nodes behind a cgnat working?
This is what I see on the node in Washington (state). The darkstat is monitoring virtual adapter where there is only node. Connection is symmetric gigabit.
Sure, past 40 or so minutes average ingress to that node was 28000 kilobytes per second, which translates to about 225 megabit/second. It uses a quarter of available gigabit bandwidth.
If you wanted to let it cook 24h or so… to see if nodes report issues (as not everybody is watching here, but would notice your-node-is-offline emails)… then that would be fine by me!
Oh, and congrats for hitting your numbers! Cash bonuses for every member of the team!
That test is planned for later maybe tomorrow. I think today we will first do the thing with excluding 20% of the nodes and also reduce the long tail to make this even more resource friendly.
The CPU usage is distributed evenly among the cores, which is nice (yeah, total CPU usage is higher than that node process - my scripts that parse logs also need resources).
I have to admit these tests are really putting my mind at rest and make me realise even a potato node can handle a lot of hammering if it’s dedicated only for Storj.
My PI5s are doing just fine.
I have a very old rack mount NAS running 3 (unvetted) nodes. It only has a Celeron J3355 and 2 GB of RAM. It’s being fairly squeezed but it’s handling over 100MBit traffic and could possibly do a bit more if I didn’t have other machines running nodes on the same IP.
i think You forgot to turn back On the ip /24 net rule,
a node that shares ip with 3-4 others is hiting his network max ingress right now ;>
and it never does
(edit okay currently 1 neighbor, so total 2 on same ip,
You made a miracle,
You chased out my unwanted neighbors:)
maybe theirs nodes crashed:P )
Edit2:
bedise that, i like how low RAM usage storagenode.exe service is using (only like 500MB, not any crazy 2000MB)
