OK. I will bring them some bombons and flowers and tell them there’s nothing to fear. 
1 Like
I… I can’t… I don’t… I don’t even know what to say… 
2 Likes
Holy! You’re bouncing off the top of a 1Gbps connection! Sweet!
2 Likes
I know, I also noticed that. Soooo jealous! I can’t get more than about 60 Mbit 
On a separate note, just noticed emails with an incident notification for US1 which fits more or less the timelines of the reduced traffic. I guess now we know 
Well, I just got around to restarting 12 of my 14 nodes in order to trigger the one time file walker after the 1.104 update to get rid of the trash miscalculation. I skipped 2 nodes since they were still running GC. So now basically all nodes are running some sort of file walker and I gotta say, I’m impressed. System is much more responsive than previously in such situations. Data coming in at decent speeds, barely any canceled transfers. Bring it on!
3 Likes
It’s time to buy new drives 
1 Like
Can confirm this as well. Even running used+gc+trash on some nodes concurrently, I’m still taking in/out traffic a lot better than before.
2 Likes
No fear!
3 Likes
All the struggling RPi nodes out there now…
3 Likes
That’s crazy! Is that like 200 nodes getting hit with the performance test?
I accept this defeat. But only to upgrade and come back)
4 Likes
- But not everyone accepts data and some share IP addresses
1 Like
That’s still amazing. Not only do I not own the drives to pull that off… I don’t have the Internet connection… or ability to get that many IPs… or even the compute to run that many containers! 
It’s really damn expensive
1 Like
it’s like 16Mbps per node average. I remember old times when Stefan could push 80-90Mbps sustained per node with wild 150-200Mbps peaks 
It is a bit too early for these kind of numbers. We haven’t hit our goal yet. Scaling up the satellite DB at the moment.
4 Likes
Lovely to see my nodes chomping down all that data without breaking a sweat, even with most of them still running filewalkers or GC. Disk writes even after RAM and SSD cache saturation seem much more efficient this way. So far I’m very happy.
5 Likes