It was just an example really but 7 mins is what I used to do on test nets when nodes would be very busy receiving records for the first minute or so. So I wanted a pause of a couple of mins. And I wanted a number that isn’t divisible by 2 because it seemed there was a peak in a node’s activity every 2 mins or so. So this would hopefully smooth out the peaks.
1 Like
Right, thanks. So, what was the problem there? Was it the CPU usage? Or bandwidth?
1 Like
Yes, Task Manager - Services Tab
Not sure what mechanism is used within the manager, but it sits there showing no sign of life for a very long time…when doing it manually is 1 minute and a safenode-manager stop afterwards confirms state.
1 Like
This is the registry refresh. I’m working on this problem just now. It is actually doing something, it is just taking a long time. Anyway, that will be removed.
4 Likes
It was both really. But mainly bandwidth usage. This was on the testnets when nodes would receive hundreds of records as soon as they started. So maybe it’s not as necessary at the moment because when a node starts it is only busy for a few seconds. I still think a delay is a good idea though, especially on something like a Pi4.
1 Like
Yeah, but I would hope that it wouldn’t have to be as long as 7 minutes.
1 Like
Less than a minute would do it just now. But back then it needed to be more than a minute. And I wanted to avoid a multiple of 2 mins because of what looks like a predictable peak in the communication that a node does.
Just stopped all but 1 node. Turns out the peaks in activity are a lot less regular than I thought. But there’s still a bit of pattern. And it’s very spiky.
These are the timings between peaks for the last few mins.
1’ 45"
2’ 59"
3’ 00"
1’ 34"
1’ 29"
2’ 55"
2’ 20"
0’ 42"
2’ 58"
1’ 43"
1’ 07"
2’ 58"
So there is always a peak every approximately 3 mins. And maybe one around 1’ 30" to 1’ 45"
I think I’ll just stop messing around and start nodes at 1 min intervals!
It would be nice if there were a way of smoothing the bandwidth usage for nodes though! If it were more smooth the peaks in site bandwidth used wouldn’t be as bad when the peaks from different nodes coincide and more nodes could be run.
1 Like