v0.10.9.x Help test RC2 forking issues

Status
Not open for further replies.

Lariondos

Well-known Member
Foundation Member
Apr 8, 2014
89
61
158
For me its normal variance - have a look at http://tdrk.poolhash.org/blocks/masterpay.txt to get a impression how the payouts are distributed.

Code:
[...]
last 90 blocks
     18 mpgNHgxhtGHhYX98wam8X9DWzpNMNaNuEA
      9 n36fd8g9ad9bph8zcmVbLPXYbkqr5H59zF
      9 muj9oHwCtoWLcrALSfV7yQS2yh1KLpbXo6
      8 mpmb1TWTTAvZNiwU91GAX1numoZeLusN8N
      7 mtLYCQ6it9eDjqGurjqBmtFSzebK8ttyaL
      7 mkLN57jC9sYqMkp8RBcJF2H9hvhz6vabwQ
      5 mzCKvrXb42KekMJcFGhneNVnJRcuRPc5Mt
      5 mxv11kABs5n2aLWHytKgFQvESd2f9vwywL
      4 muNRbrwBSgTCdHTYL5R2Eokvj9H1q21ytC
      4 mpd9So5Ryybp5zgxssDxArS8k8z4ie2i26
      4 mfpSMCgkVjrtkSVhfswSWUsdbRUxU6gFFm
      2 msBm7XEsX4x5DSkLcC44tayXS3hYyN8wBR
      2 ms9JqmHQ6QrLbeFFyAp22cofqnR5PNuHkK
      2 mpKNkHEdgtxQAjTCkGB4ZwHAED74qwjDvj
      2 mnX22j3tEyovhif1GeecQ8YCAYzWHuXEv3
      1 mw23256bSTqEkP6ThTRrY3KNicB9cfwzLR
      1 -
Lets say there were 576 blocks paid in the last 24 hours, and 15 active masternodes, then the expected median for one node is 38.4 payouts - which matches the average payouts for your two nodes quite nicely: (55 + 25) / 2 = 40

So yes: It's variance and the timeframe of 1 day is to small to judge wether there is a significant deviation. The distribution algorithm is using (random) tx hashes for each node, to spoof it you will have to find a way to manipulate the tx-hashes .
Thank you flare for this detailed explanation. To have you here ist a tremendous advantage for all of us.
 
  • Like
Reactions: flare

flare

Administrator
Dash Core Team
Moderator
May 18, 2014
2,287
2,406
1,183
Germany
Will bring 15mh/s to testnet too. Just need to know where to point to!
Why not pointing it at your own client? - 15mh/s will get you about 68% of testnets network hash currently - i am putting 2mh/s at testnet, and get 25% of all blocks mined :D

To reproduce the intermittent forks we saw in mainnet, we need to distribute the hashpower to more miners to increase the probability of simultanous generation of blocks - nevertheless it will be very difficult to reproduce this without tweaked clients...
 
Last edited by a moderator:

yidakee

Well-known Member
Foundation Member
Apr 16, 2014
1,812
1,168
283
Why not pointing it at your own miner - 15mh/s will get you about 68% of mainnets network hash currently - i am putting 2mh/s at testnet, and get 25% of all blocks mined :D

To reproduce the intermittent forks we saw in mainnet, we need to distribute the hashpower to more miners to increase the probability of simultanous generation of blocks.
OH BOY!! now If could only pull a 51% I would own ALL TESTNET coins ... mmmmu-hahahahaha (Evil cartoon laugh).... for about what, 1 or two blocks, before getting all orphaned...

Seriously now, on friday, mini forks were created because blocks were "simultaneously" found and voted (with different votes) so they went through before network consensus. If I am bringing in so much power, this will most likely not happen as statistically the chances of cracking a block at the same time someone else does, is greatly reduced. Right?

So maybe its not a good idea to bring in so much? I've got 5x 280x under win8.1 I know there is a way to fire up sgminer/cgminer to only use one card, but not sure how. I could spread out each card (solo, p2p, nomp, etc), but I'd need help doing this (how to config)
 
Last edited by a moderator:

flare

Administrator
Dash Core Team
Moderator
May 18, 2014
2,287
2,406
1,183
Germany
If I am bringing in so much power, this will most likely not happen as statistically the chances of cracking a block at the same time someone else does, is greatly reduced. Right?
Yep, that's the challenge in testnet. If we accumulate to much hashpower in one pool, we are reducing the probabilty to get simultanous blocks - same when we distribute the power to much...

So maybe its not a good idea to bring in so much? I've got 5x 280x under win8.1 I know there is a way to fire up sgminer/cgminer to only use one card, but not sure how. I could spread out each card (solo, p2p, nomp, etc), but I'd need help doing this (how to config)
I don't know about cgminer as i am using cudaminer/ccminer, but i can start my miner and assign only a specific card to it. It seems there is a '-d' parameter in sgminer, so maybe starting with '-d 0' will only start up card 0, and '-d 1' will only start up card 1.
 

jimbit

Well-known Member
Foundation Member
May 23, 2014
228
103
203
Why not pointing it at your own client? - 15mh/s will get you about 68% of testnets network hash currently - i am putting 2mh/s at testnet, and get 25% of all blocks mined :D

To reproduce the intermittent forks we saw in mainnet, we need to distribute the hashpower to more miners to increase the probability of simultanous generation of blocks - nevertheless it will be very difficult to reproduce this without tweaked clients...
I just disabled 2 of the 3 gpus... I was getting a block every few minutes.
I can turn them back on quickly.... I can also pretty easily add 40mh to testnet for short periods of time.. just have to sync to testnet on each rig.
 
  • Like
Reactions: yidakee and flare

yidakee

Well-known Member
Foundation Member
Apr 16, 2014
1,812
1,168
283
Ah, yes I've seen the "-d 0" config before. Will look into it and test - thanks flare!

chaeplin, it would be very cool if you could do your magic and give us a simple hash distribution chart. We could have an official list of testnet pools, so we could coordinate as best as possible to distribute hash, and replicate mainnet phenomenon...?
 

jimbit

Well-known Member
Foundation Member
May 23, 2014
228
103
203
OH BOY!! now If could only pull a 51% I would own ALL TESTNET coins ... mmmmu-hahahahaha (Evil cartoon laugh).... for about what, 1 or two blocks, before getting all orphaned...

Seriously now, on friday, mini forks were created because blocks were "simultaneously" found and voted (with different votes) so they went through before network consensus. If I am bringing in so much power, this will most likely not happen as statistically the chances of cracking a block at the same time someone else does, is greatly reduced. Right?

So maybe its not a good idea to bring in so much? I've got 5x 280x under win8.1 I know there is a way to fire up sgminer/cgminer to only use one card, but not sure how. I could spread out each card (solo, p2p, nomp, etc), but I'd need help doing this (how to config)
You can easily tell cgminer what 'device' to use.. . also, after you start you can go into the console menu and disable or enable specific gpus.

in console. select 'g' then 'd' then 'deviceNumber'
 

Lariondos

Well-known Member
Foundation Member
Apr 8, 2014
89
61
158
OH BOY!! now If could only pull a 51% I would own ALL TESTNET coins ... mmmmu-hahahahaha (Evil cartoon laugh).... for about what, 1 or two blocks, before getting all orphaned...

Seriously now, on friday, mini forks were created because blocks were "simultaneously" found and voted (with different votes) so they went through before network consensus. If I am bringing in so much power, this will most likely not happen as statistically the chances of cracking a block at the same time someone else does, is greatly reduced. Right?

So maybe its not a good idea to bring in so much? I've got 5x 280x under win8.1 I know there is a way to fire up sgminer/cgminer to only use one card, but not sure how. I could spread out each card (solo, p2p, nomp, etc), but I'd need help doing this (how to config)
Couldn't you simply use the LOAD BALANCE - strategy of cgminer? The hashrate should then be equally distributed to the configured pools. Never tested it though. I use ccminer with 750ti cards. Very power efficient.
 

yidakee

Well-known Member
Foundation Member
Apr 16, 2014
1,812
1,168
283
I was under the impression that Load Balance will "pool hop", not distribute the hash power to the config'ed pools. Like 50% pool A and 50% pool b = 100% hash power to pool A, but only 50% of the time.

If this is correct, its not what we're looking for. We need as much exact distribution.
 

Lariondos

Well-known Member
Foundation Member
Apr 8, 2014
89
61
158
I was under the impression that Load Balance will "pool hop", not distribute the hash power to the config'ed pools. Like 50% pool A and 50% pool b = 100% hash power to pool A, but only 50% of the time.

If this is correct, its not what we're looking for. We need as much exact distribution.
No, what you describe is the ROTATE-strategy.
 
  • Like
Reactions: yidakee

yidakee

Well-known Member
Foundation Member
Apr 16, 2014
1,812
1,168
283
Got it !! Load Balance is the definition, but the actual parameter in config file is Quota!

https://github.com/sgminer-dev/sgminer

thanks guys! - I hope chaeplin reads my suggestion! It would be cool to coordinate the maximum spread of hash power possible.
 

chaeplin

Active Member
Core Developer
Mar 29, 2014
749
356
133
Oh miner hopping.
I do this
Code:
-o stratum+tcp://127.0.0.1:3333 -u moQD1AbtbDgPhDCXhK5x9rkUQrTHK3ZE7H -p 1 \
-o stratum+tcp://54.183.73.24:3333 -u moQD1AbtbDgPhDCXhK5x9rkUQrTHK3ZE7H -p 1 \
--rotate 2 \
2 min rotate.

If GPU is two, -d and two scripts.
 
  • Like
Reactions: GreyGhost and flare

jimbit

Well-known Member
Foundation Member
May 23, 2014
228
103
203
I will just stay solo.. getting plenty of blocks, with one gpu... let me know if we need to really step it up.. with 12 or 24 hours notice I can put 40Mh on testnet
[11:36:48] Accepted 035ff399 Diff 0.296/0.242 BLOCK! GPU 2
[11:36:49] Network diff set to 0
[11:36:49] New block detected on network
[11:41:02] Found block for Pool 0!
[11:41:03] Accepted 2e9fd02f Diff 5.491/0.241 BLOCK! GPU 2
[11:41:03] Network diff set to 0
[11:41:03] New block detected on network
[11:50:03] Network diff set to 0
[11:50:03] New block detected on network
[11:50:07] Found block for Pool 0!
[11:50:07] Accepted 0289ac2b Diff 0.394/0.190 BLOCK! GPU 2
 

pinestabe

New Member
May 1, 2014
26
9
3
Seems like it would be best if mostly everyone solomined with just 1 cpu core, that way the hashrate would be spread evenly so every tester would have an equal chance of hitting a block. This would also allow usefully stacking many mining-daemons on multicore machines. Maybe blocktimes could be reduced to help cause problems?
 

daaarkcoins

Member
May 21, 2014
95
40
68
Seems like it would be best if mostly everyone solomined with just 1 cpu core
this!

We don't need high hashrate but rather a lot of miners with similar hashrates. All I have available is an i7 and when there are people with 2 digit MHs on the network there is no point in running a miner.
 

TanteStefana

Grizzled Member
Foundation Member
Mar 9, 2014
2,861
1,854
1,283
it's a great idea but the problem is sometimes people on testnet set their stuff up, and forget about it, which might cause someone to be running a bigger rig. Then you have to find out who it is and get them to shut it off, LOL. But otherwise, excellent idea
 

jimbit

Well-known Member
Foundation Member
May 23, 2014
228
103
203
so. just checking testnet... most of my solo blocks are getting orphaned....
getinfo

{
"version" : 100811,
"protocolversion" : 70015,
"walletversion" : 60000,
"balance" : 32887.73663900,
"blocks" : 23969,
"timeoffset" : 18,
"connections" : 14,
"proxy" : "",
"difficulty" : 0.19923874,
"testnet" : true,
"keypoololdest" : 1403442813,
"keypoolsize" : 101,
"paytxfee" : 0.00000000,
"mininput" : 0.00001000,
"errors" : "Hard Fork Alert: Masternode Payments begin on Friday. Please update! http://goo.gl/ucp4m7"
}


{
"account" : "",
"address" : "muy8tLwweShthtctc7A4ojHwcvzEi8ohde",
"category" : "orphan",
"amount" : 152.00000000,
"confirmations" : 0,
"generated" : true,
"txid" : "d8c9bea011c374d15546657bec096fd8681542054bb51c46ea40f3ad77a1055f",
"time" : 1403477543,
"timereceived" : 1403477554
},
{
"account" : "",
"address" : "mvY1Kr8gxCmssB9aTM3xakR8BMXVniNkvg",
"category" : "orphan",
"amount" : 136.80000000,
"confirmations" : 0,
"generated" : true,
"txid" : "c2d07a926e045a33192325a3d124c7b7688e51c425af0060f1a454e793c206eb",
"time" : 1403478147,
"timereceived" : 1403478159
},
 

flare

Administrator
Dash Core Team
Moderator
May 18, 2014
2,287
2,406
1,183
Germany
so. just checking testnet... most of my solo blocks are getting orphaned....
Works as designed - your client (0.10.8.11) is outdated and emitting invalid blocks (re. masternode payments) - update to 0.10.10.2 and you should see different results
 

jimbit

Well-known Member
Foundation Member
May 23, 2014
228
103
203
I am pretty sure I am getting orphans because my solo rig works on a fork. Looks like we will be able to reproduce the issue. I think.
 

flare

Administrator
Dash Core Team
Moderator
May 18, 2014
2,287
2,406
1,183
Germany
I am pretty sure I am getting orphans because my solo rig works on a fork. Looks like we will be able to reproduce the issue. I think.
Orphaned blocks were not the issue of RC3 launch - actually they were expected. Outdated miners are supposed to create local forks, but as long as they do not get the majority of hashpower in the network these forks will get orphaned.
 
  • Like
Reactions: derk

jimbit

Well-known Member
Foundation Member
May 23, 2014
228
103
203
Orphaned blocks were not the issue of RC3 launch - actually they were expected. Outdated miners are supposed to create local forks, but as long as they do not get the majority of hashpower in the network these forks will get orphaned.
so versioni: v0.10.8.11-unk-beta should create a local fork?
 

flare

Administrator
Dash Core Team
Moderator
May 18, 2014
2,287
2,406
1,183
Germany
so versioni: v0.10.8.11-unk-beta should create a local fork?
Yep, that's why every pool owner had to update to 0.10.10.1 before the hard fork. Any pool/miner not updating will get orphans - that's basically the idea of a hard fork :)
 

jimbit

Well-known Member
Foundation Member
May 23, 2014
228
103
203
Yep, that's why every pool owner had to update to 0.10.10.1 before the hard fork. Any pool/miner not updating will get orphans - that's basically the idea of a hard fork :)
thxs flare. so much for my solo... actually one of my blocks did get accepted somehow.. but about 100 did not.

edit: I mined the following with the older version and it WAS accepted by the network.

Status: 153 confirmations
Date: 6/22/2014 15:54
Source: Generated
Credit: 152.00 DRK
Net amount: +152.00 DRK
Transaction ID: 20923f1842b1c9f4661ae0d9d6b5f690c64b43e51f3ed0bf540986f0a674f6b7
 
Last edited by a moderator:

flare

Administrator
Dash Core Team
Moderator
May 18, 2014
2,287
2,406
1,183
Germany
thxs flare. so much for my solo... actually one of my blocks did get accepted somehow.. but about 100 did not.
Check if the accepted block had a masternode payment - the only blocks accepted by network from old miners, are the ones which have no masternode payee. They will be confirmed by the rest of the network. The other 100 blocks you generated contained invalid payout amount, since the payout algorithm in 0.10.8.x was flawed and was fixed in 0.10.10.x
 

jimbit

Well-known Member
Foundation Member
May 23, 2014
228
103
203
We definitely have a fork going on the testnet right now.. my MN has different blockhashes then my solo rig.. my last hour of blocks have not been orphaned yet.
 
  • Like
Reactions: flare

jimbit

Well-known Member
Foundation Member
May 23, 2014
228
103
203
working on setting up another rig.. downloading the chain from the rogue solo and trying to make my chain the good one.. wth :)

real chain on 24056

fork on 24054
 

jimbit

Well-known Member
Foundation Member
May 23, 2014
228
103
203
I was unsuccessful.... all of my hours worth of blocks have been orphaned.... proper chain wins.

after a -reindex soloing on the proper chain now... found block 24061 with hash 00000001183fc6478c1ace79a4e9b1d65072b1685f748717b6320a93b5174c11

if anyone is interested in seeing if we can get many clients each running about the same hash rate,
I am going to leave 2 diifferent solo clients each with one GPU running. each at about 2.48Mh
 
Last edited by a moderator:

yidakee

Well-known Member
Foundation Member
Apr 16, 2014
1,812
1,168
283
Good job guys! Still trying to figure out how to config my miner so each card hashes on its own
 
Status
Not open for further replies.