Incredibly low success rate

I have success rate below 3% on my node.

My connection works like a charm, 300/25mbit, with 6-7ms ping to from machine I use for my node.
Node is connected to router with cable, iperf shows perfect 1gbit to other LAN machine.
I don’t have spare few TB size SSD around, I use WD Blue 5TB, disk is tested, nice 150MB/s transfer rate, connected directly to SATA port.
Node hardware is on Odroid HC2, which even has its own ‘how to install storagenode’ video.
OS is linux, node data are on ext4 partition, test made ATM:
dd if=/dev/zero of=testfile bs=4k count=200000 && sync; rm testfile
200000+0 records in
200000+0 records out
819200000 bytes (819 MB, 781 MiB) copied, 5.25743 s, 156 MB/s

Are most SNOs closer to satellites/clients than me?
Check this, all ingress traffic vs europe-north-1 satellite
What I think is that 90%+ of my ingress traffic in ‘good times’ was from europe-north-1 satellite. From 2 weeks I don’t get data from it.
My status (health) is OK for all satellites, “Status: OK >> Audit[0.0% DQ|0.0% Susp]” as script shows. Node joined all satellites 2020-06-17.

if you aren’t running v 1.9.5 then you can ignore the upload successrate
they are not accurate at all

when v1.9.5 is intalled the logs should give you accurate successrates…
its expected to hit linux and docker in like a week’s time


Welcome to the forum @Dunc4n1d4h0!

As @SGC correctly mentioned there is an issue where successful uploads are shown as upload canceled in the log. A fix for it is released in version 1.9.5 which will correctly report uploads using successrate script.

1 Like

(Offtopic: You did a sequential write test, which isn’t the only thing that counts. I hate to ask but is it an SMR disk?)

My success rate is always in the “incredibly low” range. It makes very little difference.

Fortunately nope, WDC_WD50EZRZ, as /dev/disk/by-id/ says :wink:

(Offtopic of an offtopic: it should be fairly simple to find a set of parameters for fio that would simulate how a storage node works, simulating both the blob storage and sqlite at the same time. Might be a nice exercise to learn fio.)

1 Like

FYI, we also update the famous successrate thread where the first report from 1.9.5 is in which looks very promising.

Anyway, to check performance difference I just ordered Seagate IronWolf 8TB, (ST8000VN004). Lets hope for bigger, better node :slight_smile:

i usually just copy node blobs folders… :smiley:
seems to simulate it pretty well… and it’s heavy work lol
atleast for my setup… i’m sure its different from system to system…