Hello there - a quick question - over the past months my setups has been growing. I have noticed that the nodes i started 2 months ago was vetted fast - and the new ones are not vetted yet - but have decent data.
Here is some facts i know:
Vetting process was recently changed to increase the time it takes.
Unvetted = 1% of total traffic goes to unvetted nodes.
My experience:
My unvetted nodes is taking longer than expected.
My unvetted nodes seems to get much less data than the vetted ones (though same subnet neigbours)
Questions:
Is someone else seeing unvetted nodes getting less data atm. than vetted nodes? (asking because its a new system, and it might be running slow for other reasons - low ram, old cpu).
What is the new vetting time look like?
The main goal of vetting system is to prove the stability of the new nodes for a longer period of time.
Required number of audits are bumped as we have more audits than earlier. To avoid too quick vetting, we also bumped the required number of audits with an estimation to be vetted in 3-4 weeks.
This will be adjusted more precisely with this patch (can require either a number of audit or a minimum period of time):
94c347e64 N - satellite/reputation: check node age in addition to number of audits to determine vetted status (2 weeks ago) <Jennifer Johnson> <Jennifer Li Johnson>
The other bug what we had that we gave an unfair amount of traffic to unvetted nodes.
It’s fixed here:
* 3d47568e3 N - satellite/nodeselection: unvetted selector should use threshold as a max (2 weeks ago) <Márton Elek> <Elek, Márton>
Thanks for the info, @elek . Since I’m reporting vetting progress in my earnings calculator. I’d like to incorporate this info in the progress percentage. Can you confirm you are using the release defaults of 100 audits and a minimum of 504h?
Maybe a fixed number of audits isn’t usefull, only the “number of passed audits/total number of audits” in a set period of time, like 30 days, passed audits meaning audits with OK response.
This way you only set a timeframe and you send as many audits as you want. Only the percent of them that were passed will give the node a thumbs up or down, meaning a score that should be above a limit.
You could go further with this score and present it to the SNO in real time, on the dashboard, giving him an ideea of how well the node performs and how his changes affect the node.
That’s literally what the audit and suspension scores already are.
I think the fixed number is still there just in case for some reason new nodes don’t get at least 100 audits in 21 days. You still want a representative sample, even if that means it takes a little longer.
I think the 21 days (504h) is the correct number, but I have seen some satellites take a little longer on a few new nodes I had. up to maybe 25 days? I don’t know whether there is some process that runs infrequently or there could be a backlog somewhere, or my assumption was wrong in my previous post. Would still appreciate a confirmation, @elek. If you have a moment to spare. Is it indeed the case that production satellites use 504h and that the node needs both at least 100 successful audits and an age over 504h?
You need at least 1000/350/200/200 (US1/EU1/AP1/SLC) audit and 504 hours elapsed.
100 was bumped because we didn’t have the time based settings what is implemented since that time. Now that we have it, likely the amount based audit will be decreased (but as we have more audits, 1000 audit can be done in less than 21 hours. Shouldn’t make any big difference)