iPollo G1 rejected shares

tromp · February 18, 2021, 9:52am

Running my mini miner overnight, I found this morning that out of 1130 shares submitted, 59 were rejected due to

Verification Error: endpoints don’t match up

That’s a rejection rate of over 5%, where I see no reason there should be even a single rejection. The relevant source code is at

github.com

mimblewimble/grin/blob/e7bbda81a0a815994955ce8fc71f710cd17a11bb/servers/src/mining/stratumserver.rs#L361-L503


	fn handle_submit(
		&self,
		params: Option<Value>,
		worker_id: usize,
	) -> Result<(Value, bool), RpcError> {
		// Validate parameters
		let params: SubmitParams = parse_params(params)?;
		let state = self.current_state.read();
		// Find the correct version of the block to match this header
		let b: Option<&Block> = state.current_block_versions.get(params.job_id as usize);
		if params.height != state.current_block_versions.last().unwrap().header.height
			|| b.is_none()
		{
			// Return error status
			error!(
					"(Server ID: {}) Share at height {}, edge_bits {}, nonce {}, job_id {} submitted too late",
					self.id, params.height, params.edge_bits, params.nonce, params.job_id,
				);
			self.workers.update_stats(worker_id, |ws| ws.num_stale += 1);

This file has been truncated. show original

If the stratum protocol included the siphash keys that define a Graph, then it would be trivial to pinpoint the blame, as the solution could be directly checked against that. Currently, however we need to derive them from the header that was used when sending out the job.

If the logic for looking up the header against which to verify the submitted solution were at fault, then we would see similar rejection rates for all solo-mining GPU based miners as well. Does anyone have data on that?

grin001 · February 18, 2021, 1:01pm

Do you mean that the rejection rate of ipollo mining machine is very high? If the rejection rate is about 2%, it belongs to the normal range.

Anynomous · February 19, 2021, 10:04am

When not mining solo via f2pool, I also get around 1.6% rejected shares with the G1 Mini which is similar to what I get with my RTX2080Ti, around 2% rejected shares. No successful payouts though so as soon as I figure out how, I will switch pools

Anynomous · February 26, 2021, 1:38pm

Not sure if it is in any way related, but on Grinmint I noticed that it appears like I have high percentage of invalid shares. Hard to give a precice number, but it is higher than 2% and higher than on f2pool. Also in general the graph rate observed by Grinmint appears to be lower than on f2pool. I gave my second G1 Mini to a friend, otherwise it would be nice to do a side by side comparison of mining on both pools to investigate if this is just my imagination or maybe some technical issue.
=> Problem is with the miner, f2pool does however appear to show (and possibly pay out) the hash rate based on both valid and invalids shares

xiaojay · May 1, 2021, 12:23pm

@tromp check and test this pr

github.com/mimblewimble/grin

make the job_id of every task in stratum server unique to reduce iPollo G1 rejected shares.

mimblewimble:master ← xiaojay:master

opened 12:22PM - 01 May 21 UTC

xiaojay

+4 -2

Problem to fix https://forum.grin.mw/t/ipollo-g1-rejected-shares/8432 I wo…rk with iPollo G1 miner's developer and found that the reason of ipollo miner rejected shares is grin stratum server sometimes send job with same job_id but is not same pre_work, which make miner failed to work on the new job then submit the wrong share, which to be rejected. In this pr, we use (current_state.current_block_versions.len() as u64) * bh.height as unique job_id. we have tested it in goblinpool.com, and reduce iPollo G1 rejected shares from ~5% -> ~0.2%