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

[trouserless]

Not sure why your stales are so high.  First thing it always starts at 0% that is pretty much meaningless.  It spiking to 112% in a short period of time is equally meaningless.  When you have a small number of shares the "luck" of a single stale can affect that stat by a lot.  Say hypothetically you have 10 shares w/ 1 stale (10% stale or ~100% efficiency)  The 11th share is coming in.  If it is stale your efficiency will drop to 91% if it is valid then your efficiency will rise to 102%.  What matters if the value after some time (say 12 to 24 hours @ 3 GH/s).   Once you have hundreds of shares the value of each one is less material and the number will stabilize.

makes sense, thanks.

A couple of ideas:
1) Use latest version of cgminer.  2.2.7 may be fine but some versions are less p2pool friendly.

using BAMT on usb sticks.  I can install it in a parallel directory I suppose.  Probably a good idea to keep closer to the latest release anyhow

2) Are your stales mostly dead or orphaned?  Some dead is unavoidable (similar to any other pool) but a high dead may indicate you are pushing the cards too hard and causing invalid hashes.

I've got 7 orphaned, 4 dead since earlier today (is that statistically significant?).  I'm not seeing any hardware errors reported by cgminer

3) Make sure your system running bitcoind & p2pool isn't overloaded.  Preferably this would be a non-mining system which is mostly idle.  bitcoind & p2pool use some memory and their operations are very atency sensitive.  If they are slow to issue LP you are going to have excessive stales.

bitcoind is running on what appears to be an underpowered server (see earlier response to Forrest).  P2pool software is piggy-backing on a miner with 2GB and seems to be not taxing at all (that miner is a BAMT USB stick FWIW).

4) Make sure p2pool and bitcoind have a high number of connections (this means port forwarding)

check - iptables rules forwarding inbound 9333 to p2pool server and verified - 21 peers, 11 incoming

5) Make sure your internet connection is solid w/ low latency and good uptime.

Verizon FIOS - 20/20...I'm very lucky to have it in my area.

As a test try using public p2pool node and see if your stale rates improve.  If it does then it is something in your internet, LAN, p2pool machine which is being inefficient.  If it doesn't improve then the issue lies with the miner.

That's a great idea...going to do this tomorrow morning, leave it for a day or so.  Thank you!

[DeathAndTaxes]

using BAMT on usb sticks.  I can install it in a parallel directory I suppose.  Probably a good idea to keep closer to the latest release anyhow

Here is why I usually do.  install new version in /opt/miners folder under a new directory.  test it on one rig by stopping gpumon and running "new" cgminer manually from that directory.  If everything looks good rename the cgminer folder to cgminer-2.2.7 and rename new cgminer folder (i.e. cgminer-2.3.1 to cgminer.  start from gpumon.  Makes sure everything looks good for a day and then upgrade all other rigs.

It has never failed me.  BAMT has no knowledge of what version of cgminer it is running it simply executes the cgminer in the /opt/miners/cgminer folder.

bitcoind is running on what appears to be an underpowered server (see earlier response to Forrest).  P2pool software is piggy-backing on a miner with 2GB and seems to be not taxing at all (that miner is a BAMT USB stick FWIW).

Stop the troubleshooting.  This is it then.  bitcoind needs to be responsive.  I mean fraction of a second responsive.  Once you get it running fast somewhere put p2pool there too then point all your rig(s) at that p2pool instance.  Having bitcoind and p2pool on the same "server" means less latency.

I think you will find one bitcoind is fast and responsive your stales will plummet.

[echris1]

I am running bitcoind and p2pool on a sempron 145 right now, but will soon have another rig up and running with a dual core processor.  I will probably move my p2pool over to that rig then (or run it on both as a backup) then I'll be able to give people a real world example of processor powers impact on p2pool stales. 

It might be worth checking if your single core can be unlocked, I know some of the semprons can be, mine can't unfortunately, but worth a look.

[c_k]

Bad idea unlocking core imo

[echris1]

I suppose, but we are talking about a 40 dollar processor, so its not a huge loss if something goes wrong =)

[echris1]

So, I might be wrong, but I just did a git pull earlier and now I keep seeing this every now and then:

Code:

  Pool: 267GH/s Stale rate: 8.4% Expected time to block: 6.7 hours
2012-03-08 02:52:25.538244 Incoming connection to peer 77.91.201.77:33987 established. p2pool version: 2 '5ad7c89'
2012-03-08 02:52:25.539173 Got share hash, requesting! Hash: 86a8a3f4
2012-03-08 02:52:25.724369 Sending 1 shares to 77.91.201.77:33987
2012-03-08 02:52:25.725745 RECV shares 0100fd2801010000008c428aa19c5e7994dbb352b430a4e506a526cecf7fe47958f304000000000000e84453d66984eb29f05f6b94dfce3fd4144b3b8944d9545a621a8cd2e32e7b965263584f0c350b1ad480175e99b93670dcb09552a12756c4d581e2...
2012-03-08 02:52:25.727150 > Error handling message: (see RECV line)
2012-03-08 02:52:25.727829 > Traceback (most recent call last):
2012-03-08 02:52:25.728164 >   File "/usr/lib/python2.7/dist-packages/twisted/internet/selectreactor.py", line 146, in _doReadOrWrite
2012-03-08 02:52:25.728636 >     why = getattr(selectable, method)()
2012-03-08 02:52:25.729543 >   File "/usr/lib/python2.7/dist-packages/twisted/internet/tcp.py", line 460, in doRead
2012-03-08 02:52:25.730067 >     rval = self.protocol.dataReceived(data)
2012-03-08 02:52:25.730413 >   File "/home/miner/p2pool/p2pool/p2p.py", line 55, in new_dataReceived
2012-03-08 02:52:25.730968 >     old_dataReceived(data)
2012-03-08 02:52:25.731593 >   File "/home/miner/p2pool/p2pool/util/datachunker.py", line 40, in _DataChunker
2012-03-08 02:52:25.731945 >     wants = receiver.send(buf.get(wants))
2012-03-08 02:52:25.732330 > --- <exception caught here> ---
2012-03-08 02:52:25.732657 >   File "/home/miner/p2pool/p2pool/bitcoin/p2p.py", line 53, in dataReceiver
2012-03-08 02:52:25.733014 >     self.packetReceived(command, type_.unpack(payload))
2012-03-08 02:52:25.733376 >   File "/home/miner/p2pool/p2pool/p2p.py", line 68, in packetReceived
2012-03-08 02:52:25.733734 >     bitcoin_p2p.BaseProtocol.packetReceived(self, command, payload2)
2012-03-08 02:52:25.734063 >   File "/home/miner/p2pool/p2pool/bitcoin/p2p.py", line 66, in packetReceived
2012-03-08 02:52:25.734418 >     handler(**payload2)
2012-03-08 02:52:25.734744 >   File "/home/miner/p2pool/p2pool/p2p.py", line 199, in handle_shares
2012-03-08 02:52:25.735096 >     self.node.handle_shares([p2pool_data.Share.from_share(share, self.node.net, self) for share in shares], self)
2012-03-08 02:52:25.735476 >   File "/home/miner/p2pool/p2pool/data.py", line 187, in from_share
2012-03-08 02:52:25.735829 >     raise ValueError('unknown share type: %r' % (share['type'],))
2012-03-08 02:52:25.736208 > exceptions.ValueError: unknown share type: 0
2012-03-08 02:52:25.737452 Lost peer 77.91.201.77:33987 - Connection was closed cleanly.

:19.499847  Shares: 0 (0 orphan, 0 dead) Stale rate: ??? Efficiency: ??? Current payout: 0.0413 BTC
2012-03-08 03:25:19.499969  Pool: 272GH/s Stale rate: 9.5% Expected time to block: 6.5 hours
2012-03-08 03:25:20.160618 Incoming connection to peer 67.183.247.222:60716 established. p2pool version: 2 '053d7ca'
2012-03-08 03:25:20.161605 Got share hash, requesting! Hash: 1a6c8e71
2012-03-08 03:25:20.266125 Sending 1 shares to 67.183.247.222:60716
2012-03-08 03:25:20.372467 RECV shares 0100fd88010100000067a2a2950ec6bad48b6c37088f5e9531f725fa730b3a49059e0500000000000004a3f8996d8b4b95e03f5d57f68625e0bef73a8edbefba292612bc7feb5b9305fe4d584f0c350b1ab73d4a27dd55c17161a4cb5868763ed263aaad...
2012-03-08 03:25:20.374046 > Error handling message: (see RECV line)
2012-03-08 03:25:20.374423 > Traceback (most recent call last):
2012-03-08 03:25:20.374775 >   File "/usr/lib/python2.7/dist-packages/twisted/internet/selectreactor.py", line 146, in _doReadOrWrite
2012-03-08 03:25:20.375150 >     why = getattr(selectable, method)()
2012-03-08 03:25:20.375504 >   File "/usr/lib/python2.7/dist-packages/twisted/internet/tcp.py", line 460, in doRead
2012-03-08 03:25:20.375845 >     rval = self.protocol.dataReceived(data)
2012-03-08 03:25:20.376244 >   File "/home/miner/p2pool/p2pool/p2p.py", line 55, in new_dataReceived
2012-03-08 03:25:20.376582 >     old_dataReceived(data)
2012-03-08 03:25:20.376904 >   File "/home/miner/p2pool/p2pool/util/datachunker.py", line 40, in _DataChunker
2012-03-08 03:25:20.378011 >     wants = receiver.send(buf.get(wants))
2012-03-08 03:25:20.378355 > --- <exception caught here> ---
2012-03-08 03:25:20.378689 >   File "/home/miner/p2pool/p2pool/bitcoin/p2p.py", line 53, in dataReceiver
2012-03-08 03:25:20.379071 >     self.packetReceived(command, type_.unpack(payload))
2012-03-08 03:25:20.379417 >   File "/home/miner/p2pool/p2pool/p2p.py", line 68, in packetReceived
2012-03-08 03:25:20.379792 >     bitcoin_p2p.BaseProtocol.packetReceived(self, command, payload2)
2012-03-08 03:25:20.380138 >   File "/home/miner/p2pool/p2pool/bitcoin/p2p.py", line 66, in packetReceived
2012-03-08 03:25:20.380540 >     handler(**payload2)
2012-03-08 03:25:20.380842 >   File "/home/miner/p2pool/p2pool/p2p.py", line 199, in handle_shares
2012-03-08 03:25:20.381192 >     self.node.handle_shares([p2pool_data.Share.from_share(share, self.node.net, self) for share in shares], self)
2012-03-08 03:25:20.381564 >   File "/home/miner/p2pool/p2pool/data.py", line 187, in from_share
2012-03-08 03:25:20.381911 >     raise ValueError('unknown share type: %r' % (share['type'],))
2012-03-08 03:25:20.382259 > exceptions.ValueError: unknown share type: 0
2012-03-08 03:25:20.383555 Lost peer 67.183.247.222:60716 - Connection was closed cleanly.

Is this a share that I am submitting that is somehow returning an error message?

Edited:

Just got this again, looks like its everytime I am finding a share, any ideas?  Is the bitcoin 6rc now required, because I am still using 0.5.2

Well, after digging in the logs a bit it looks like this happens whether or not I'm mining.

[broken]

It cant manage to upgrade to bitcoin-0.6.0-rc2. It always seems it still starts rc1.

# The files are apperently different

broken@rig02:~/bitcoin$ ls -la bitcoin-0.6.0rc1-linux/bin/64/ | grep bitcoin
-rwxr-xr-x 1 broken broken 4247696 2012-02-07 19:46 bitcoind
-rwxr-xr-x 1 broken broken 8891013 2012-02-07 19:46 bitcoin-qt

broken@rig02:~/bitcoin$ ls -la bitcoin-0.6.0rc2-linux/bin/64/ | grep bitcoin
-rwxr-xr-x 1 broken broken 4251824 2012-02-29 21:59 bitcoind
-rwxr-xr-x 1 broken broken 9053450 2012-02-29 21:59 bitcoin-qt

# Bitcoin is stopped
broken@rig02:~/bitcoin$ ps -ef | grep coin
broken     2165     1  0 Feb14 ?        02:40:52 ./namecoind -daemon

# Starting bitcoin
broken@rig02:~/bitcoin$ cd bitcoin-0.6.0rc2-linux/bin/64/
broken@rig02:~/bitcoin/bitcoin-0.6.0rc2-linux/bin/64$ ./bitcoind -daemon
bitcoin server starting
************************
EXCEPTION: 11DbException      
Db::put: Cannot allocate memory      
bitcoin in ProcessMessage()

# Bitcoin is now running
broken@rig02:~/bitcoin$ ps -ef | grep coin
broken     2165     1  0 Feb14 ?        02:40:53 ./namecoind -daemon
broken    15663     1 10 09:45 ?        00:00:12 ./bitcoind -daemon

# Its the correct binary
broken@rig02:~/bitcoin$ readlink -f /proc/15663/exe
/home/broken/bitcoin/bitcoin-0.6.0rc2-linux/bin/64/bitcoind

# Bitcoin gives error and is stuck at 170059
broken@rig02:~/bitcoin/bitcoin-0.6.0rc2-linux/bin/64$ ./bitcoind getinfo
{
    "version" : 60000,
    "protocolversion" : 60000,
    "balance" : 0.00000000,
    "blocks" : 170059,
    "connections" : 8,
    "proxy" : "",
    "difficulty" : 1496978.59502557,
    "testnet" : false,
    "keypoololdest" : 1322670639,
    "keypoolsize" : 103,
    "paytxfee" : 0.00000000,
    "errors" : "WARNING: Displayed transactions may not be correct!  You may need to upgrade, or other nodes may need to upgrade."
}

Anyone have a idea how to upgrade?

[tucenaber]

I have the same issue. I started getting

WARNING: Displayed transactions may not be correct!  You may need to upgrade, or other nodes may need to upgrade.

with every RCP query and tried to upgrade to rc2 without success.

bitcoind sits out database error once in a while:

************************
EXCEPTION: 11DbException       
Db::get: Cannot allocate memory       
bitcoin in ProcessMessage()       

Now 0.5.2 gives the same error.

The only difference from the above post is that my transaction count is 170061

I have tried to delete the database log.0000* but that doesn't change anything.

(I am running Arch linux which use a different version of BDB, could that be an issue? I haven't had any issues with that in the past)

[stevegee58]

tucenaber and broker, you both need to upgrade to bitcoin 0.6.0rc2.

If you're on windows, look at my most recent post in this thread.  The 0.6.0rc2 installer doesn't work right.  You have to use the zip file.
I was sure I'd upgraded to rc2 but because the installer didn't work I was still on rc1 and had the problem you had.

[broken]

tucenaber and broker, you both need to upgrade to bitcoin 0.6.0rc2.

If you're on windows, look at my most recent post in this thread.  The 0.6.0rc2 installer doesn't work right.  You have to use the zip file.
I was sure I'd upgraded to rc2 but because the installer didn't work I was still on rc1 and had the problem you had.

I'm pretty sure I'm running bitcoin 0.6.0rc2. But i guess rc1 broke my bitcoin db somehow.
Im currently testing on another machine, where I switched directly from 0.5.2 to 0.6.0rc2 at block 167000. But still have to download 1000 blocks. If this works ill just copy the database to the other machine.

edit:

Does p2pool still work with 0.5.2 or is 0.6.0rc2 absolutely necessary?
I knew I shouldn't have upgraded to a unstable version on a productive system.

[kano]

...
Does p2pool still work with 0.5.2 or is 0.6.0rc2 absolutely necessary?
I knew I shouldn't have upgraded to a unstable version on a productive system.

Yeah they are called 'rcX' for a reason.
Nothing should be depending on them.

[twmz]

Does p2pool still work with 0.5.2 or is 0.6.0rc2 absolutely necessary?

Yes, you can still use 0.5.x

[DeathAndTaxes]

I'm pretty sure I'm running bitcoin 0.6.0rc2. But i guess rc1 broke my bitcoin db somehow.
Im currently testing on another machine, where I switched directly from 0.5.2 to 0.6.0rc2 at block 167000. But still have to download 1000 blocks. If this works ill just copy the database to the other machine.

1) You can use 0.5 bitcoind.

2) If you used the windows installer you almost certainly are not running 0.6.0 rc2.  It will say it installed but it doesn't overwrite everything from 0.6.0 rc1.

3) For p2pool I recommend using a bitcoind instance w/ a "blank wallet" and manually setting the payment address.  That way a bad wallet simply means delete wallet.dat and relaunch bitcoind to create a new one.

4) For custom windows installs the zip file is preferable.  You can drop it in a custom folder, delete the GUI bitcoin, and set data directory to the same folder.  Upgrading or replacing simple means deleting the folder.

[ChanceCoats123]

Bad idea unlocking core imo

Why would that be a bad idea? From my experience with unlocking Amd chips, it either works and it's 100% stable, or it doesn't load into Windows/Linux/Whatever-the-hell. So there is no possible damage to be done by trying to unlock. My miner runs on a Gigabyte UD5P with a 555 unlocked to 4x and it works just fine.

[broken]

I'm pretty sure I'm running bitcoin 0.6.0rc2. But i guess rc1 broke my bitcoin db somehow.
Im currently testing on another machine, where I switched directly from 0.5.2 to 0.6.0rc2 at block 167000. But still have to download 1000 blocks. If this works ill just copy the database to the other machine.

1) You can use 0.5 bitcoind.

2) If you used the windows installer you almost certainly are not running 0.6.0 rc2.  It will say it installed but it doesn't overwrite everything from 0.6.0 rc1.

3) For p2pool I recommend using a bitcoind instance w/ a "blank wallet" and manually setting the payment address.  That way a bad wallet simply means delete wallet.dat and relaunch bitcoind to create a new one.

4) For custom windows installs the zip file is preferable.  You can drop it in a custom folder, delete the GUI bitcoin, and set data directory to the same folder.  Upgrading or replacing simple means deleting the folder.

I'm using linux and I'm pretty sure I'm running the correct binary as you can see in my earlier post:

broken@rig02:~/bitcoin$ ps -ef | grep coin
broken     2165     1  0 Feb14 ?        02:40:53 ./namecoind -daemon
broken    15663     1 10 09:45 ?        00:00:12 ./bitcoind -daemon

broken@rig02:~/bitcoin$ readlink -f /proc/15663/exe
/home/broken/bitcoin/bitcoin-0.6.0rc2-linux/bin/64/bitcoind

You are correct, it is best to use a empty wallet on that machine and that is what I do.
I tried to delete everything except the blockchain, but still the same problem.
I also tried to run 0.5.2, but its stuck at 170059, too.

broken@rig02:~/bitcoin/bitcoin-0.5.2-linux/bin/64$ ./bitcoind getinfo
{
    "version" : 50200,
    "balance" : 0.00000000,
    "blocks" : 170059,
    "connections" : 8,
    "proxy" : "",
    "generate" : false,
    "genproclimit" : -1,
    "difficulty" : 1496978.59502557,
    "hashespersec" : 0,
    "testnet" : false,
    "keypoololdest" : 1331211591,
    "keypoolsize" : 101,
    "paytxfee" : 0.00000000,
    "errors" : "WARNING: Displayed transactions may not be correct!  You may need to upgrade, or other nodes may need to upgrade."
}

I think 0.6.0rc1 broke my blk0001.dat,  blkindex.dat or something else not located in .bitcoin/ (is there something else?)
When I get home I will also delete these files. (I have a 170190+ blockchain here that I will try first, before downloading the whole thing again)

[cabin]

I think 0.6.0rc1 broke my blk0001.dat,  blkindex.dat or something else not located in .bitcoin/ (is there something else?)
When I get home I will also delete these files. (I have a 170190+ blockchain here that I will try first, before downloading the whole thing again)

I thought this too.. but mine did recover after ~5 min after installing rc2. Also for those of you wondering about the bitcoind response time, I believe these lines of code will print out when there is trouble in main.py arond line 152:

        def set_real_work1():
            my_time = time.time()
            work = yield getwork(bitcoind)
            my_time2 = time.time() - my_time
            if (my_time2 > 0.2):
                print "slow bitcoind: %.02f" % (my_time2,)


Mine are usually under 0.2 seconds, but on occasion take up to 2 seconds, which is interesting.

[Frizz23]

How come our "reference" address http://blockexplorer.com/address/1Kz5QaUPDtKrj5SqW5tFkn7WZh8LmQaQi4 gets almost constant bounty per block (always ~0.16 per block) - while mine varies so much (0.02 - 0.17) http://blockexplorer.com/address/1EZMh5fH33TFWLh6enwL86gam5LnEE2o4v ?

I had my machine running 24 hours the lase couple of days.

[DeathAndTaxes]

How come our "reference" address http://blockexplorer.com/address/1Kz5QaUPDtKrj5SqW5tFkn7WZh8LmQaQi4 gets almost constant bounty per block (always ~0.16 per block) - while mine varies so much (0.02 - 0.17) http://blockexplorer.com/address/1EZMh5fH33TFWLh6enwL86gam5LnEE2o4v ?

I had my machine running 24 hours the lase couple of days.

Variance.

If donation is a small % of a large amount of GH/s then the payment will be small but constant.

0.16 per block would be roughly 100% of 1.15 GH/s.  However the donation is most likely the default for most users that donate.  So it isn't 100% of 1.15 GH/s.  It is 0.5% of ~200 GH/s.  If you had 200 GH/s your payout per block would be very constant too.

[broken]

How come our "reference" address http://blockexplorer.com/address/1Kz5QaUPDtKrj5SqW5tFkn7WZh8LmQaQi4 gets almost constant bounty per block (always ~0.16 per block) - while mine varies so much (0.02 - 0.17) http://blockexplorer.com/address/1EZMh5fH33TFWLh6enwL86gam5LnEE2o4v ?

I had my machine running 24 hours the lase couple of days.

Isn't 1Kz5QaUPDtKrj5SqW5tFkn7WZh8LmQaQi4 forrestv's author donation address, which always gets 0,5% of 50BTC for each block. Some people disabled donation so its a bit less, but pretty stable.

[broken]

I think 0.6.0rc1 broke my blk0001.dat,  blkindex.dat or something else not located in .bitcoin/ (is there something else?)
When I get home I will also delete these files. (I have a 170190+ blockchain here that I will try first, before downloading the whole thing again)

I thought this too.. but mine did recover after ~5 min after installing rc2.

I gave it another hour, but it didn't recover. I copied the blockchain files from another machine and now it works.