Open Joseba73 opened 3 years ago
I have also noticed this with other combos of miner or algo. I think I've had it happen with autolykos2 on T-Rex as the rejecting algo. But the T-Rex changelog said some stuff about corrupt shares and I thought it was just that version. It happens so randomly I'm not sure if the problem is still there with v0.24.2... But it's not just when switching to ethash.
My suspicion is just something with miner init didn't work right -- shutdown of old miner to startup of new miner needs settle-delay to allow for the previous miner to fully exit? I have not seen this with a service restart in systemd with miners, it either goes some hundreds of ms slower or actually awaits all the killed pids to go away before starting again.
Definitely some sort of state/timing problem. Adding a relaunch delay in general is probably good, it really forks crazy fast when a GPU goes absent, etc.
This issue was monitored since a few days. When switching from cuckaroo29b to ethash mining all shares are rejected as duplicates. Then of course IP is banned. I think I didn't noticed it before because this switching is quite rare but this might be a new issue, dunno. All other monitored switching pairs work fine from weeks. The result is the same with GMiner or T-rex with ethash: duplicate shares
...