Been playing around with my Pool 0, 1, and 2 since I never fixed that read-only issue with my RPI...
Just wondering if making the same node for 0 and 1 would alleviate some of the "5 minute failover shenanigans" Kano mentions.
My logic is that if a node has a quick disconnect-reconnect and it's enough to cause a failover, that may be just enough time to see it as available again after the failover, and it ends up using the same node. Otherwise, it'll just resort to the main us.kano.is node.
Any thoughts?
I wouldn't rely on the failover in the miner code working well - I regularly hear from people asking questions that mean that it doesn't.
Make the 2nd one simply just another kano node somewhere else so if it fails over it will be to a node a bit further away.
The millisecond numbers people see are pretty much irrelevant unless they're well over 100ms - which should really only be the other side of the planet.
Pools in general used to take many times longer than that to switch work unless they were SPV pools mining empty blocks.
So it really isn't a big deal to take 100ms or a bit more.
Of course work change here has always been much quicker than that
There used to be stats about that info about pools, that I wrote, in the cgminer API so you could actually check, but ck broke all that long ago and never fixed it
They're still there ('stats') but they don't show the pool specific timing info any more - though there's still useful info in there
?
