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

[ThiagoCMC]

This fact that is not possibile to tell exactly what was found and when is a big problem, IMHO, which could stop miners from switching to p2pool.

+1

p2pool claims to be totally transparent. Users having full control (as in opposite to big mining pools).

Yet we don't even know for sure when / what was found.

Here's a great command to list found blocks:

Code:

bitcoind listtransactions

+1

[Proofer]

Here's a great command to list found blocks:

Code:

bitcoind listtransactions

For Linux and/or bash users:  here's a bash function I wrote yesterday -- it omits the txid lines from listtransactions, but more important (to me) it adds a human-readable time to the time of each one:

Code:

btx () {
    bitcoind listtransactions $* | while read name colon value
        do
            if [ $colon ] ; then
                if [ $name != '"txid"' ] ; then # omit txid line
                    if [ $name = '"time"' ] ; then  # provide a readable version of time
                        echo '"time"' : $value = `date -d "1970-01-01 $value seconds UTC" +"%Y-%m-%d %T %z"`
                    else
                        echo "$name : $value"
                    fi
                fi
            else
                echo $name
            fi
        done
}

Thus, this...

Code:

        "txid" : "aca39f64e4f71f984b9693fe1dbd6dc9361162d94db10b1b814121cf3da44a9f",
        "time" : 1328894325

is replaced by....

Code:

"time" : 1328894325 = 2012-02-10 09:18:45 -0800

(The UTC offset is specific to your local system's time zone setting.)

[Ente]

Still using GuiMiner, it throws:

2012-02-10 17:50:49: Miner "p2pool": 127.0.0.1:9332 10/02/2012 17:50:49, long poll: new block 000004f709b08c9c
2012-02-10 17:50:53: Miner "p2pool": 127.0.0.1:9332 10/02/2012 17:50:53, Unexpected error:
2012-02-10 17:51:03: Miner "p2pool": 127.0.0.1:9332 10/02/2012 17:51:03, Unexpected error:
2012-02-10 17:51:09: Miner "p2pool": 127.0.0.1:9332 10/02/2012 17:51:09, Unexpected error:
2012-02-10 17:51:10: Miner "p2pool": 127.0.0.1:9332 10/02/2012 17:51:10, long poll: new block 000004f709b08c9c
2012-02-10 17:51:16: Miner "p2pool": 127.0.0.1:9332 10/02/2012 17:51:16, long poll: new block 000004f709b08c9c
2012-02-10 17:51:18: Miner "p2pool": 127.0.0.1:9332 10/02/2012 17:51:18, long poll: new block 000004f709b08c9c

Any ideas?

Do all of the rigs have identical configurations? It looks like you're using poclbm as the backend.. Maybe try downloading the newest version of poclbm from https://github.com/m0mchil/poclbm/downloads and dropping it into GUIMiner's directory?

Now thats strange..
After rebooting all rigs, now all (0.8.1) p2pool instances seem to give/get work to guiminer. The rig with 0.8.3 is the only one with "unexpected error" with guiminer.
All 8 rigs have the same versions, configs, tools, everything is identical (except 0.8.1/0.8.3)

I try to get more info from poclbm/guiminer..

Ente

[Wandering Albatross]

P2Pool should listen on port 9332 on all interfaces. Is this a problem..? You mentioned just needing to connect external miners, which many people already do.

It's not a problem for me now, I modified main.py. I tried to add the diff to an issue for p2pool on github. I'm pretty sure that my diff got garbled on github. sorryboutthat I'm not githubish yet.

I have tested it and it does work.

[Wandering Albatross]

Here's a great command to list found blocks:

Code:

bitcoind listtransactions

It would be cool if bitcoind had a side-channel interface so that queries could be made while it was running.
Probably via http or ssh.

[rjk]

Here's a great command to list found blocks:

Code:

bitcoind listtransactions

It would be cool if bitcoind had a side-channel interface so that queries could be made while it was running.
Probably via http or ssh.

It does, just start a new instance in another command window with your desired arguments. They will be forwarded to the running bitcoind and the values returned to you.

[Tittiez]

I've been playing around with making a p2pool recent stats page.  It's not done yet, and the backend needs a lot of work, but it's done enough to give you an idea:

http://btcstats.net/p2pool/

• Includes Recent pool hash rate and stale rate
• Recent blocks are shown on the chart as vertical lines so that you can "see" variance
• Recent blocks are also listed in a table with links to the details

That's really cool. Can we integrate a luck graph into there too'?

[forrestv]

After rebooting all rigs, now all (0.8.1) p2pool instances seem to give/get work to guiminer. The rig with 0.8.3 is the only one with "unexpected error" with guiminer.
All 8 rigs have the same versions, configs, tools, everything is identical (except 0.8.1/0.8.3)

Ah, thanks, that gave me enough information to find the (likely) problem - this commit: https://github.com/forrestv/p2pool/commit/d39081784f6 !

GUIminer doesn't use a real JSON parser, and that additional field might be confusing it.. Which backend are you using?

By the way, those releases are really old. There have been some important fixes since then.. Most people just use something recent from git master. I might get back into the habit of making releases soon, though.

[Ente]

Do you run P2Pool on each individual rig? If so, any reason why?

Yep, every rig has its own p2pool node.
My friend prefers to not be dependent on one machine's p2pool.

Talking about statistics: the probability that one p2pool from any of the rigs crashes is higher than if there is just one, (locally) centralized p2pool where all rigs connect to.. I'll talk him out of that next week :-)

I rolled back to 0.8.1 and that miner works now again.
The guiminer log only said "unexpected error" too.

I will have a second look next week, and maybe switch to cgminer etc anyway, and hopefully to one p2pool node where all rigs connect to too.

Ente

[Ente]

Ah, thanks, that gave me enough information to find the (likely) problem - this commit: https://github.com/forrestv/p2pool/commit/d39081784f6 !

GUIminer doesn't use a real JSON parser, and that additional field might be confusing it.. Which backend are you using?

By the way, those releases are really old. There have been some important fixes since then.. Most people just use something recent from git master. I might get back into the habit of making releases soon, though.

Guiminer 2011-06-14
with poclbm "-f 100 -w 128 -v"
I see no obvious way to find out poclbm's version. If you wish to know that I have a second look.

edit:
Oh, I assumed both gid and the tar'ed version would be the same.. Good to know, I'll get the gid version then!
Maybe drop a note on that on the downloadpage? ;-)

Thank you for your help and all the time and energy you put in this!

Ente

[forrestv]

Ente: Here's a version with the submitold commit reverted: https://github.com/forrestv/p2pool/tarball/nosubmitold

Do you think you could test it to see if it fixes the problem?

EDIT: Also, does GUIMiner have any log files? There should be an additional message after the "Unexpected error" lines, but it's not there. Maybe it's appearing on the terminal or something?

[kano]

I've been playing around with making a p2pool recent stats page.  It's not done yet, and the backend needs a lot of work, but it's done enough to give you an idea:

http://btcstats.net/p2pool/

• Includes Recent pool hash rate and stale rate
• Recent blocks are shown on the chart as vertical lines so that you can "see" variance
• Recent blocks are also listed in a table with links to the details

Hmm nice.
I presume the lack of history can be overcome with help from forrestv?

[Wandering Albatross]

It does, just start a new instance in another command window with your desired arguments. They will be forwarded to the running bitcoind and the values returned to you.

Then it is cool. Works great.

[twmz]

Hmm nice.
I presume the lack of history can be overcome with help from forrestv?

Yes, he gave me a database of old hashrate stats.  I just have to spend time getting it into my database.  We don't have old state rate data.  I'm not sure if stale rate over time is all that interesting though, so maybe it isn't important and should be replaced with something like # of active users over time.

I also want to look at luck, but I will admit that I don't yet know how I am going to calculate that with only the information I have.  But forrest's graph does it using this same data, so I probably just need to chat with him more and see what approach he is using.

I really need to also get a real backend for it because right now it is using the rrd files that the standard p2pool uses.  I'd like to move this to a real database on my web host instead of using the rrd files.

But tomorrow is my wife's birthday, so I don't know how much time I can reasonably spend on this over the weekend.  I'll probably squeeze a few hours in.

[kano]

No hurry
Was just thinking out loud - but you already had thought of it anyway.

[ThiagoCMC]

Upgrading P2Pool node... 

Guys, when is the best time to upgrade P2Pool node?

I mean, is there any announcements of new P2Pool versions?! If yes, where?!

Now I have:

Code:

~/p2pool/run_p2pool.py --version
becb233

But, is this the version you guys are using?!

Thanks!
Thiago

[miscreanity]

Code:

~/p2pool/run_p2pool.py --version
becb233

The current git pull is:
aa520c1

[twmz]

Ok, historical data from forrest is loaded.  It's interesting to see how quickly things have grown relative to how long the pool floated steadily at around 10-15 GH/s.

http://btcstats.net/p2pool/

• Includes Recent pool hash rate and stale rate active users
• Recent blocks are shown on the chart as vertical lines so that you can "see" variance
• Recent blocks are also listed in a table with links to the details

[Ente]

Ente: Here's a version with the submitold commit reverted: https://github.com/forrestv/p2pool/tarball/nosubmitold

Do you think you could test it to see if it fixes the problem?

EDIT: Also, does GUIMiner have any log files? There should be an additional message after the "Unexpected error" lines, but it's not there. Maybe it's appearing on the terminal or something?

Good morning,

I tested the "new old" version you linked. It works, connects faster and all, but the (guiminer) error still persists.
Still, the guiminer "console" says "unexpected error:", and thats all it says in its logfile too (in guiminer.log and guiminer.exe.log).

I'll experiment a bit with connecting (different) guiminer instances to (different) p2pool nodes to find out which one causes trouble here..

Ente

[Ente]

It gets more and more strange..

Rig "Miner04" uses forrestv-p2pool-0.8.4-1-gadba786
Rig "Miner03" uses forrestv-p2pool-release-0.8.1-0-gc7feb00

Rig "Miner04" uses Guiminer v2011-06-14
Rig "Miner03" uses Guiminer v2011-06-14

Guiminer "Miner03" --> p2pool "Miner03": finds "minerblocks" and "shareblocks"
Guiminer "Miner04" --> p2pool "Miner04": drops "unexpected error:" in guiminer
Guiminer "Miner03" --> p2pool "Miner04": finds "minerblocks" with 100% in (), finds "shareblocks"
Guiminer "Miner04" --> p2pool "Miner03": finds "minerblocks" with 50% in (), finds no "shareblocks"

So obviously its some problem on Guiminer's (or poclbm) side on "Miner04".

In Guiminer you see for example:

Accepted: 471 (256)
Stale: 0 (0)

I didnt find out what the number in () means, yet.

All other rigs are like "Miner03" and behave well.

Bah, I'm losing my patience with this windows stuff..

Ente

edits underlined