I would have expected some lost races due to long tail cancellations. However, we may be looking at the better performing nodes here. People who post right after the update are likely to be better performing nodes. Additionally, transfers may finish so fast that they are done before they can be cancelled. It’s possible more than the success threshold sticks around if that happens.
Once I get the update on my docker node I’ll post my score. Fortunately for this test, I have the opposite of a better performing node. My success rate is currently 9.786%. If my new numbers with the update are above 95% I will be very suspicious. As for transfers finishing to fast, I might be able to do some traffic shaping to test that.
I just got the update. It looks like I will need to disable max-concurrent-requests for this new version of the software. I’m getting a lot of:
ERROR piecestore upload rejected, too many requests {"live requests": 101, "requestLimit": 100}.
I would not have thought that a limit of 100 was too small but it seems to keep the requests open for a while and they are adding up over time. The rejections are messing up my “Acceptance Rate” score along with the success rate score.
Quick question here - why do you assume that we may look at the better performing nodes? Don’t necessarily see the connection between people who post after update = have a better rig?
It wasn’t an assumption, but rather speculation. Enthusiasts tend to have enthusiast hardware. Doesn’t seem so far fetched to me. And it seems that the post right after yours confirmed that it’s not a universal thing to have 99+%.
Ok clear - yeah I get that, even though I would say there are enough raspberry 3 owners (now maybe 4) who just keep it running with an additional disk … Just thinking as my feeling was overall we have more ‘geeks’ in here running a node which could fit that image.
Nevertheless so far it looks pretty good in terms of upload. Once my docker updates I hope it’ll look similar. In my case ‘unfortunately’ I also changed my connection to a better one (hence old successrate improved from 25% to 40% already). Would have been a great test to see how my bad connection plays out here but this one is gone in a week (had both in parallel for a whlie)
Yep, just to confirm, both my nodes are RPi 4’s with USB 3.0 connected 5400rpm HDDs (WD Elements). Although, last week I did get a new ISP with 1Gbit fiber up/down, so that might be one reason my successrate numbers pretty high on this new version. Back when I was on V1.6.4, after the ISP upgrade, my successrates were closer to ~10-14% for uploads…with this new version, I’m still rocking >99% upload successrate on both nodes after 8-12 hours.