[1500 TH] p2pool: Decentralized, DoS-resistant, Hop-Proof pool - page 422.

Mr. Jinx

member

Activity: 112

Merit: 10

Quote from: paniczklos on February 26, 2014, 01:29:28 AM

Hi guys, is there a way to check why I'm not getting fees from every mined block (and the total fees are much lower than the set 1%) ?
Right now avarage payout of two biggest miners only is ~4000BC, but I'm not getting any fees ;/ (It's not about greed - I just don't understand it).
Hashrate is constant (and growing).

I had the same question in another topic:

Quote from: roy7 on February 02, 2014, 09:02:13 AM

Quote from: Mr. Jinx on February 02, 2014, 08:36:15 AM

Should you get a 1% fee everytime the miners are paid? I haven't received any fee yet, while the active miners have received their share payments.
Or will this fee be paid on a daily base? I got the feeling something is setup wrong.

It isn't intuitive, but each share a miner submits to you has a 1% chance of becoming your share instead. So you'll see a lot of variance in your pool fee income, but long-term it should even out (at the expense of higher variance for the miner as well).

Maybe one day p2pool will allow the fee amount to be recorded in the share, like the donation amount is, so it can be paid exactly when each share is paid and all shares remain with the miner.

After a few weeks I can confirm this behavior. Sometimes there are days without a fee, and then suddenly you get a few in a row. Looking back a month it seems to be about 1%!

paniczklos

full member

Activity: 322

Merit: 101

Hi guys, is there a way to check why I'm not getting fees from every mined block (and the total fees are much lower than the set 1%) ?
Right now avarage payout of two biggest miners only is ~4000BC, but I'm not getting any fees ;/ (It's not about greed - I just don't understand it).
Hashrate is constant (and growing).

http://freebtc.eu:8336/static/graphs.html?Day

VarDiff

full member

Activity: 196

Merit: 100

P2Pool added to mining pools database
Good luck, miners.

phillipsjk

legendary

Activity: 1008

Merit: 1001

Let the chips fall where they may.

Well I think I found out where the extra 6 seconds of lag came from:

Code:

2014-02-25 13:53:22.835376 Punishing share for 'Block-stale detected! height(6f2f28d90af36171b9bc41e7986c7bb24943c37bec4ef054) < height(151b94b7ca0744b491b3fcbf950435984bc36c5a97c9a08e9) or
19015f53 != 19015f53'! Jumping from 1ce9c6ad to 212471a6!
...
2014-02-25 13:53:24.455802 New work for worker! Difficulty: 1.000000 Share difficulty: 529462.142225 Total block value: 25.027840 BTC including 106 transactions
.
.
.
2014-02-25 14:05:10.552019 Transaction db65517b5b4b7314e86aa52a1f9e0ea3be37044eab1a73e408ea07f74ec5e383 rescued from peer latency cache!
...
2014-02-25 14:05:17.463766 New work for worker! Difficulty: 1.000000 Share difficulty: 553972.015873 Total block value: 25.105139 BTC including 529 transactions
.
.
.
2014-02-25 14:14:57.135562 P2Pool: 17337 shares in chain (9424 verified/17341 total) Peers: 6 (0 incoming)
...
2014-02-25 14:15:00.183437 P2Pool: 17338 shares in chain (9425 verified/17342 total) Peers: 6 (0 incoming)
...
2014-02-25 14:15:05.433806 New work for worker! Difficulty: 1.000000 Share difficulty: 565095.416800 Total block value: 25.057434 BTC including 304 transactions

I may be misinterpreting things, but I appears to take my machine 1-7 seconds to put work together.

IYFTech

hero member

Activity: 686

Merit: 500

WANTED: Active dev to fix & re-write p2pool in C

Exactly.

If you do find a proxy that is compatible with p2pool without a daft DOA rate I'm sure there are a few hundred miners who would like to hear about it...... Wink

phillipsjk

legendary

Activity: 1008

Merit: 1001

Let the chips fall where they may.

Quote from: IYFTech on February 23, 2014, 07:03:01 PM

Quote from: phillipsjk on February 23, 2014, 06:18:59 PM

Edit: Since AFAIK, the getwork protocol does not allow workers to be interrupted with new work, we should be able to estimate the expected stales given a specific block frequency. If we assume a 13s worst case latency, that works out to at most 43% stales with a 30 second target. If we assume a 6.5s average, that works out to 21.7% stale. -- that does seem high.

Edit: Apparently Longpolling works around HTTP limitations by having the miner request new work immediately. The server then does not respond until new work is ready. Testing time.

See this:

https://bitcointalksearch.org/topic/m.4556966

I don't know of anyone who has got slush's proxy to work with p2pool, I've tried countless times myself - it just doesn't work, in fact I have a feeling that that's what is now causing my p2pool start up error/warning. Your only choice with a blade is to use +1 at the end of your user name/addy which, tbh, is a bit of a waste..... Wink

OK, I finally got P2Pool running and learned the truth. One thing that confused me is that the +1 option does not appear to be documented anywhere. What it does is tell the blade to report diff-1 shares. Apparently, the blades do not support higher difficulty.

Test results are in: no longpoll support. 21% stale for me, I guess. (confirmed with netstat -- only transient connections show up on port 9332) The reported DOA rate varies between 20-40% for some reason.

PS: Slush's proxy does not work because P2Pool does not appear to support the stratum protocol. I may try a different proxy if I think it will result in less CPU usage (for example, using compiled, rather than interpreted code). Better power savings would result from replacing the CPU though.

Edit: P2Pool does support Stratum. Slush's proxy not working is more complex. P2Pool tries to automatically detect whether the worker is using stratum or not. While at the same time, the slush proxy tries to automatically detect if the pool is using Stratum or not (using the getwork protocol). I tried removing that detection from the slush code, and it still didn't work.

PatMan

hero member

Activity: 924

Merit: 1000

Watch out for the "Neg-Rep-Dogie-Police".....

Quote from: kano on February 24, 2014, 05:39:04 PM

Hmm ... so how many TH/s is the 'development team' being paid?

There's a dev team?........ Cheesy

roy7

sr. member

Activity: 434

Merit: 250

Quote from: kano on February 24, 2014, 05:39:04 PM

Hmm ... so how many TH/s is the 'development team' being paid?

Since donations are stored on the share chain, you could actually whip up a tool to calculate that if you were curious enough.

(I realize you might be asking a rhetorical question in response to IYFTech.)

kano

legendary

Activity: 4634

Merit: 1851

Linux since 1997 RedHat 4

Hmm ... so how many TH/s is the 'development team' being paid?

IYFTech

hero member

Activity: 686

Merit: 500

WANTED: Active dev to fix & re-write p2pool in C

This is an ongoing age old problem, scroll back & you'll see.

Not enough development, not enough (lately almost zero) support, not enough compatibility, not user friendly enough......the list goes on & on.

It's a shame.

Tegija

member

Activity: 112

Merit: 10

Just Fun!

Quote from: jedimstr on February 24, 2014, 03:39:38 PM

So on the BTC side, P2Pool has been having horrible luck lately... and the Hashrate on the pool has essentially been flat or declining since the new year.

Anything we can do to encourage more people to use P2Pool (I saw a PR push on Reddit around the holidays and the GHash.io panic on Reddit), or to enhance our chances of getting a few blocks? It seems every other block these days is GHash.io blocks now.

it would be nice if we could start some campaign to get more people to p2pool. i am trying hard and have found some new people, but considering that at this time every 2-3 days there come more new TH to the network than whole p2pool totally has, we need really good strategy to get new miner to p2pool.

it is pretty amazing that people do not understand that it is for everybody good if p2pool gets bigger and more healthy.

i hope it will change soon an we have more people mining with us.

jedimstr

hero member

Activity: 798

Merit: 1000

So on the BTC side, P2Pool has been having horrible luck lately... and the Hashrate on the pool has essentially been flat or declining since the new year.

Anything we can do to encourage more people to use P2Pool (I saw a PR push on Reddit around the holidays and the GHash.io panic on Reddit), or to enhance our chances of getting a few blocks? It seems every other block these days is GHash.io blocks now and the previous outrage has died down since BTCGuild and Eligius have caught up and split the top 3 spots: https://blockchain.info/pools

IYFTech

hero member

Activity: 686

Merit: 500

WANTED: Active dev to fix & re-write p2pool in C

Quote from: Mr. Jinx on February 24, 2014, 03:18:25 AM

Quote from: forrestv on February 23, 2014, 05:19:08 PM

Mr. Jinx, occasional memory leaks have plagued P2Pool nodes for a while. Python/Twisted makes it pretty easy to leave a reference to something or leave a task running in the background... several of them have been patched, though.

Try looking through P2Pool's log file for lines that contain a ">" character, which indicates an uncaught exception/traceback. A task dying that periodically removes old shares was the cause of a previous memory leak; maybe this is something similar.

Ah, thanks. That makes it easier to troubleshoot!

This one happens alot. Always the same peers that gives this error (2).
What could be causing this?

Code:

2014-02-24 05:00:39.106919 > in handle_share_hashes:
2014-02-24 05:00:39.107328 > Traceback (most recent call last):
2014-02-24 05:00:39.107653 > Failure: twisted.internet.defer.TimeoutError: in GenericDeferrer
2014-02-24 05:00:39.109908 Lost peer x.x.x.x:36098 -
2014-02-24 05:00:39.110112 Connection was aborted locally, using
2014-02-24 05:00:39.110291 L{twisted.internet.interfaces.ITCPTransport.abortConnection}.
2014-02-24 05:00:39.110502
2014-02-24 05:00:39.110721 @since: 11.1
2014-02-24 05:00:39.110918 .[

Anyone who can shed some light on this?

This is p2pool disconnecting from older, incompatible versions of p2pool (v11.1) - it's a good thing. The others I'm not sure of, sorry.

Mr. Jinx

member

Activity: 112

Merit: 10

Quote from: forrestv on February 23, 2014, 05:19:08 PM

Mr. Jinx, occasional memory leaks have plagued P2Pool nodes for a while. Python/Twisted makes it pretty easy to leave a reference to something or leave a task running in the background... several of them have been patched, though.

Try looking through P2Pool's log file for lines that contain a ">" character, which indicates an uncaught exception/traceback. A task dying that periodically removes old shares was the cause of a previous memory leak; maybe this is something similar.

Ah, thanks. That makes it easier to troubleshoot!

This one happens alot. Always the same peers that gives this error (2).
What could be causing this?

Code:

2014-02-24 05:00:39.106919 > in handle_share_hashes:
2014-02-24 05:00:39.107328 > Traceback (most recent call last):
2014-02-24 05:00:39.107653 > Failure: twisted.internet.defer.TimeoutError: in GenericDeferrer
2014-02-24 05:00:39.109908 Lost peer x.x.x.x:36098 -
2014-02-24 05:00:39.110112     Connection was aborted locally, using
2014-02-24 05:00:39.110291     L{twisted.internet.interfaces.ITCPTransport.abortConnection}.
2014-02-24 05:00:39.110502
2014-02-24 05:00:39.110721     @since: 11.1
2014-02-24 05:00:39.110918     .[

This one a few times, but I don't think this is cause memory problems:

Code:

2014-02-24 05:05:18.137532 Worker xxxxxxxxxxx submitted share with hash > target:
2014-02-24 05:05:18.137946     Hash:   xxxxxxxxxxxxxxx
2014-02-24 05:05:18.138166     Target: xxxxxxxxxx

And this one a few times, while the pool hasn't been restarted for days:

Code:

2014-02-24 05:34:43.706844 > Couldn't link returned work's job id with its handler. This should only happen if this process was recently restarted!

Anyone who can shed some light on this?

roy7

sr. member

Activity: 434

Merit: 250

Quote from: IYFTech on February 23, 2014, 07:03:01 PM

I don't know of anyone who has got slush's proxy to work with p2pool, I've tried countless times myself - it just doesn't work, in fact I have a feeling that that's what is now causing my p2pool start up error/warning. Your only choice with a blade is to use +1 at the end of your user name/addy which, tbh, is a bit of a waste..... Wink

Provided you are running your own node and not flooding network traffic to a public node, what's wrong with using the +1 solution?

smooth

legendary

Activity: 2968

Merit: 1198

Quote from: forrestv on February 23, 2014, 07:59:57 PM

The ASICMiner blades only do getwork requests, which already contain a midstate that P2Pool generated. Using the Stratum proxy has no or very little advantage over connecting directly to P2Pool. (As far as I know; don't have hardware to test with.)

The only advantage would be bandwidth if you are using a p2pool public node. But this doesn't seem to work very well anyway.

I haven't tried blades with a local p2pool node and I've now sold them off so I can't.

forrestv

hero member

Activity: 516

Merit: 643

Quote from: IYFTech on February 23, 2014, 07:03:01 PM

Quote from: phillipsjk on February 23, 2014, 06:18:59 PM

Quote from: Hunterbunter on February 19, 2014, 05:50:05 AM

Hey guys, what bad things come from having a high getwork latency? I'm under the impression that stratum makes getwork irrelevant, but I'm a bit worried that mine is quite high (sometimes a few seconds, you can see it here http://www.blisterpool.com/stats in the graphs bit). Is this a symptom of anything in particular? The cpu doesn't seem to be under any enormous stress, but it does peak every now and then...is this the cause?

Also, I helped a miner get configured to run on my p2pool node, and he has some ASICMiner blades (~10.7GH each). Google search results (most seem to be from a year ago) have all told me that they naturally have high DOA rate with p2pool, without any real way to fix it. Is this still the case? His dead rate is around 40-50%. I also noticed the server was getting absolutely hammered with hash > target spam, and I suggested to the miner to use his bitcoin address+1 for his username. It reduced the server spam drastically (from 100/sec to several/sec), and reduced his DOA a little bit, but it also reduced his mean hashing power by about 10%. Could anyone explain to me what's going on here? I'd like to help him get better results.

Is this miner running a stratum proxy locally? each of the 32 chips seem to mostly ask for work independently.
My blade gets work 150 times per minute, meaning that the latency should be at most 12.8 seconds (150/32*60).

I don't actually have P2Pool working yet, but moved the midstate calculations to my Stratum proxy on the assumption it can calculate mid-state faster than the blade.

See this:

https://bitcointalksearch.org/topic/m.4556966

I don't know of anyone who has got slush's proxy to work with p2pool, I've tried countless times myself - it just doesn't work, in fact I have a feeling that that's what is now causing my p2pool start up error/warning. Your only choice with a blade is to use +1 at the end of your user name/addy which, tbh, is a bit of a waste..... Wink

The ASICMiner blades only do getwork requests, which already contain a midstate that P2Pool generated. Using the Stratum proxy has no or very little advantage over connecting directly to P2Pool. (As far as I know; don't have hardware to test with.)

IYFTech

hero member

Activity: 686

Merit: 500

WANTED: Active dev to fix & re-write p2pool in C

Quote from: phillipsjk on February 23, 2014, 06:18:59 PM

Quote from: Hunterbunter on February 19, 2014, 05:50:05 AM

Hey guys, what bad things come from having a high getwork latency? I'm under the impression that stratum makes getwork irrelevant, but I'm a bit worried that mine is quite high (sometimes a few seconds, you can see it here http://www.blisterpool.com/stats in the graphs bit). Is this a symptom of anything in particular? The cpu doesn't seem to be under any enormous stress, but it does peak every now and then...is this the cause?

Also, I helped a miner get configured to run on my p2pool node, and he has some ASICMiner blades (~10.7GH each). Google search results (most seem to be from a year ago) have all told me that they naturally have high DOA rate with p2pool, without any real way to fix it. Is this still the case? His dead rate is around 40-50%. I also noticed the server was getting absolutely hammered with hash > target spam, and I suggested to the miner to use his bitcoin address+1 for his username. It reduced the server spam drastically (from 100/sec to several/sec), and reduced his DOA a little bit, but it also reduced his mean hashing power by about 10%. Could anyone explain to me what's going on here? I'd like to help him get better results.

Is this miner running a stratum proxy locally? each of the 32 chips seem to mostly ask for work independently.
My blade gets work 150 times per minute, meaning that the latency should be at most 12.8 seconds (150/32*60).

I don't actually have P2Pool working yet, but moved the midstate calculations to my Stratum proxy on the assumption it can calculate mid-state faster than the blade.

See this:

https://bitcointalksearch.org/topic/m.4556966

I don't know of anyone who has got slush's proxy to work with p2pool, I've tried countless times myself - it just doesn't work, in fact I have a feeling that that's what is now causing my p2pool start up error/warning. Your only choice with a blade is to use +1 at the end of your user name/addy which, tbh, is a bit of a waste..... Wink

phillipsjk

legendary

Activity: 1008

Merit: 1001

Let the chips fall where they may.

Quote from: Hunterbunter on February 19, 2014, 05:50:05 AM

Hey guys, what bad things come from having a high getwork latency? I'm under the impression that stratum makes getwork irrelevant, but I'm a bit worried that mine is quite high (sometimes a few seconds, you can see it here http://www.blisterpool.com/stats in the graphs bit). Is this a symptom of anything in particular? The cpu doesn't seem to be under any enormous stress, but it does peak every now and then...is this the cause?

Also, I helped a miner get configured to run on my p2pool node, and he has some ASICMiner blades (~10.7GH each). Google search results (most seem to be from a year ago) have all told me that they naturally have high DOA rate with p2pool, without any real way to fix it. Is this still the case? His dead rate is around 40-50%. I also noticed the server was getting absolutely hammered with hash > target spam, and I suggested to the miner to use his bitcoin address+1 for his username. It reduced the server spam drastically (from 100/sec to several/sec), and reduced his DOA a little bit, but it also reduced his mean hashing power by about 10%. Could anyone explain to me what's going on here? I'd like to help him get better results.

Is this miner running a stratum proxy locally? each of the 32 chips seem to mostly ask for work independently.
My blade gets work 150 times per minute, meaning that the latency should be at most 12.8 seconds (150/32*60).

I don't actually have P2Pool working yet, but moved the midstate calculations to my Stratum proxy on the assumption it can calculate mid-state faster than the blade.

Edit: Since AFAIK, the getwork protocol does not allow workers to be interrupted with new work, we should be able to estimate the expected stales given a specific block frequency. If we assume a 13s worst case latency, that works out to at most 43% stales with a 30 second target. If we assume a 6.5s average, that works out to 21.7% stale. -- that does seem high.

Edit: Apparently Longpolling works around HTTP limitations by having the miner request new work immediately. The server then does not respond until new work is ready. Testing time.

roy7

sr. member

Activity: 434

Merit: 250

It's just a warning, not an error. And importing a library twice shouldn't hurt anything.

Topic: [1500 TH] p2pool: Decentralized, DoS-resistant, Hop-Proof pool - page 422. (Read 2591964 times)