When a block is found is it submitted directly to bitcoind network or does it have to be approved by nodes before?
Just seems that lately p2pool is running much worse than normal variance would suggest.
Topic: [1500 TH] p2pool: Decentralized, DoS-resistant, Hop-Proof pool - page 693. (Read 2591920 times)
Yowch! - been a few P2Pool orphans lately ...
Block# -> block internal timestamp (not the exact time the block was found)
181922 -> 2012-05-28 05:28:39 (won by Deepbit)
181892 -> 2012-05-28 00:06:07 (won by Unknown)
181621 -> 2012-05-26 05:12:31 (won by Unknown)
181492 -> 2012-05-25 04:47:58 (won by Slush)
though P2Pool did win this one:
181303 -> 2012-05-23 17:53:16 (against Unknown)
So anyway, 4 in the last 3 days
I wonder if the people who made these blocks have low connection counts, poor p2pool performance or slow network connectivity from the pool to the internet?
Of course orphans are expected every so often, but 4 in the last 3 days almost suggests something else ...
Block# -> block internal timestamp (not the exact time the block was found)
181922 -> 2012-05-28 05:28:39 (won by Deepbit)
181892 -> 2012-05-28 00:06:07 (won by Unknown)
181621 -> 2012-05-26 05:12:31 (won by Unknown)
181492 -> 2012-05-25 04:47:58 (won by Slush)
though P2Pool did win this one:
181303 -> 2012-05-23 17:53:16 (against Unknown)
So anyway, 4 in the last 3 days
I wonder if the people who made these blocks have low connection counts, poor p2pool performance or slow network connectivity from the pool to the internet?
Of course orphans are expected every so often, but 4 in the last 3 days almost suggests something else ...
As seemingly inappropriate as this is... I HATE ORPHANS! 
Fucking Deepbit
As seemingly inappropriate as this is... I HATE ORPHANS!
shit - laughed so loud i woke the neighbors!
though, yea. agreement.
As seemingly inappropriate as this is... I HATE ORPHANS!
(It dawned on me my dsl modem was probably doing it, and I was right. Once I changed it to bridged, all the sudden I'm getting incoming connections.)
Do you have a separate router after the modem? If yes, then you are supposed to set the DSL modem to bridged, otherwise you lose lots of packets.Didn't lose packets, I just couldn't forward anything from the router, because unsolicited incoming connections never got there to forward.
Quote
Bridged means you are letting anyone on the internet at anything connected to the modem.
Non-bridged means you have to tell the modem to port forward (or set a DMZ - from a security level, effectively the same as being bridged)
Non-bridged means you have to tell the modem to port forward (or set a DMZ - from a security level, effectively the same as being bridged)
When I initially got the modem years ago, I could not figure out how to get it working with bridged. Not sure why, if it was on Verizon's side, or mine. Ended up switching to DHCP, and everything worked. Aside from port forwarding, try as I might, I couldn't get that to work. Only recently did I realize it was either Verizon outright blocking everything, or my modem. Fortunately it was the latter.
M
Bridged means you are letting anyone on the internet at anything connected to the modem.
Non-bridged means you have to tell the modem to port forward (or set a DMZ - from a security level, effectively the same as being bridged)
Non-bridged means you have to tell the modem to port forward (or set a DMZ - from a security level, effectively the same as being bridged)
(It dawned on me my dsl modem was probably doing it, and I was right. Once I changed it to bridged, all the sudden I'm getting incoming connections.)
Do you have a separate router after the modem? If yes, then you are supposed to set the DSL modem to bridged, otherwise you lose lots of packets. 
try to temprarily use my p2pool and see if it still happens. I'm running on windows server 2k8 r2 x64.
My cgminer doesn't do that, even not set as fail over only.
I'm using the latest version 2.4.1, on windows 7, and windows vista. Both setup the same way, with one 7970, and both send work to the failover servers. It did when there were 4 gpus on it too. I attributed it to my local p2pool not being able to supply work fast enough. It seems to be a little better now that I have two servers, and they are forced linked together (-n), and I got port forwarding finally working so I have more than 10 connections now on my main server. (It dawned on me my dsl modem was probably doing it, and I was right. Once I changed it to bridged, all the sudden I'm getting incoming connections.) But both cgminers are still sending work to the backup. Right now the backup is reporting 222m/h, and the third backup (p2pmining.com) is reporting 62m/h.
I have about 3g/h thrown at it, which I didn't think was that much.
Only cgminer is doing it. My other miner is using phoenix 1.7.5 through guiminer.
I'm half tempted to remove the backups in cgminer to force them all to go to the main server. But that PC isn't as stable as I'd like it to be yet, I want backups.
M
My cgminer doesn't do that, even not set as fail over only.
Run cgminer as failover only
It is failover only. That's the puzzling piece. cgminer, sitting on the same PC as p2pool, sends anywhere from 5-20% of my traffic to the failover server.
M
Run cgminer as failover only
I'm now running two p2pools locally, on two different machines. I assume it's okay if both point to the same payout address? Or is that going to confuse the network?
Not easier run one p2pool and point both miners to one machine?My issue is cgminer likes sending a portion, somewhat small, but still a portion of my hashing to my backup server, which wasn't mine. So I thought why not run p2pool somewhere else.. since I already have 3 machines running all the time, two of which have up to speed bitcoind running? Seems to be working well, although I added my original backup server as #3 and cgminer still insists on sending some hash there. Much smaller than before, but it's still sending it there.
M
I'm now running two p2pools locally, on two different machines. I assume it's okay if both point to the same payout address? Or is that going to confuse the network?
Not easier run one p2pool and point both miners to one machine? 
I'm now running two p2pools locally, on two different machines. I assume it's okay if both point to the same payout address? Or is that going to confuse the network?
Yup, that is fine. You may want to use different addresses to make it easier to track them, but you don't need to.
I'm now running two p2pools locally, on two different machines. I assume it's okay if both point to the same payout address? Or is that going to confuse the network?
The problem most likely lies in the overhead of p2pool. The shares in the p2pool blockchain have data added, the current miners and shares and payouts that have to be computed in as transactions and hidden in the bitcoin blockchain as an unknown transaction as well as all the current bitcoin blockchain transactions to be considered a valid share. the 10 second longpolls also mean that many created shares are rejected as late/stale and discarded especially for slower miners.
As long as the miner can hash without having to work on something that can't become a block, overhead isn't a noticeable problem (it can increase the chances of orphan blocks, but the overhead should reach a level where latencies are around 30 seconds to have a ~5% hit on our luck). Given a 10s share interval target, a 30s latency would be detected by insane levels of dead/orphan p2pool shares.
The problem most likely lies in the overhead of p2pool. The shares in the p2pool blockchain have data added, the current miners and shares and payouts that have to be computed in as transactions and hidden in the bitcoin blockchain as an unknown transaction as well as all the current bitcoin blockchain transactions to be considered a valid share. the 10 second longpolls also mean that many created shares are rejected as late/stale and discarded especially for slower miners.
Was just thinking about our bad luck and decided to write it down just in case (might give ideas to others and make my understanding of p2pool more clear).
Could a proportion of miners not submitting stales in some peculiar way explain it ?
Are all stales both counted for the total pool hashrate and used as potential blocks if possible ?
Is it possible that some configuration (miner software + p2pool) could both make a potential-block stale share been thrown away (miner might consider it stale or p2pool) and at the same time broadcast other stales that are used to compute the total hashrate ?
We thought of an attack (were some miners could deliberately withheld blocks) but I don't see other ways that either an overvalued hashrate (?) or blocks being ignored. How is the hashrate computed anyway (I suppose you have to count stale and valid p2pool shares, are stales broadcasted to all nodes ?) ?
Could a proportion of miners not submitting stales in some peculiar way explain it ?
Are all stales both counted for the total pool hashrate and used as potential blocks if possible ?
Is it possible that some configuration (miner software + p2pool) could both make a potential-block stale share been thrown away (miner might consider it stale or p2pool) and at the same time broadcast other stales that are used to compute the total hashrate ?
We thought of an attack (were some miners could deliberately withheld blocks) but I don't see other ways that either an overvalued hashrate (?) or blocks being ignored. How is the hashrate computed anyway (I suppose you have to count stale and valid p2pool shares, are stales broadcasted to all nodes ?) ?
Jump to: