Author Topic: Mining on Linux instructions?  (Read 26815 times)

0 Members and 1 Guest are viewing this topic.

Offline FreeTrade

  • Board Moderator
  • Hero Member
  • *****
  • Posts: 700
    • View Profile
Re: Mining on Linux instructions?
« Reply #45 on: November 06, 2013, 03:31:19 am »
I just found lots of 'generated block is stale' in my debug logs.  Is this normal?  I'm assuming this means the miner found a block but right before it found it someone else found a block.  I wouldn't expect to see this very often unless blocks were comin g really fast?  But I found like 10 of these errors.

This is due to a combination of the long hash time and the fast block time currently being experienced. When the difficulty adjusts, and we're seeing 5 minute blocks, you'll see a lot fewer staler blocks.

Also I agree killing threads as soon as new blocks are found would be a good improvement and result in fewer stale blocks too.
“People should be more sophisticated? How are you gonna get that done?” - Jerry Seinfeld reply to Bill Maher

Offline bytemaster

Re: Mining on Linux instructions?
« Reply #44 on: November 06, 2013, 01:26:35 am »
So then does the current implementation have any implications for preferring less cores or anything like that?  Does mining with a higher # of cores result in more "stale blocks" for some reason?  Or would that not make a difference.

To be honest, I did not look at what FreeTrade implemented very closely.  I only verified that his results matched my reference design which had only half the HPM as FreeTrade managed to achieve.   Of course my reference design was written for clarity rather than performance and our white paper on Momentum talks about potential optimization strategies.   I am sure it won't be long until someone comes up with something better.
For the latest updates checkout my blog: http://bytemaster.bitshares.org
Anything said on these forums does not constitute an intent to create a legal obligation or contract between myself and anyone else.   These are merely my opinions and I reserve the right to change them at any time.

Offline bsamjolake

  • Newbie
  • *
  • Posts: 14
    • View Profile
Re: Mining on Linux instructions?
« Reply #43 on: November 06, 2013, 01:20:58 am »
So then does the current implementation have any implications for preferring less cores or anything like that?  Does mining with a higher # of cores result in more "stale blocks" for some reason?  Or would that not make a difference.

Offline bytemaster

Re: Mining on Linux instructions?
« Reply #42 on: November 06, 2013, 01:17:55 am »
I'm gonna go with implementation issue, considering kill -9 can stop anything dead in its tracks.  Once you see a valid block all current mining instructions should be halted and restarted with new latest block.

Yes, there is much work to be done on the algorithm implementation.  It is clearly not thread safe (Crashing all of the time) and has no ability to cancel mid-hash.
For the latest updates checkout my blog: http://bytemaster.bitshares.org
Anything said on these forums does not constitute an intent to create a legal obligation or contract between myself and anyone else.   These are merely my opinions and I reserve the right to change them at any time.

Offline bsamjolake

  • Newbie
  • *
  • Posts: 14
    • View Profile
Re: Mining on Linux instructions?
« Reply #41 on: November 06, 2013, 01:16:06 am »
I'm gonna go with implementation issue, considering kill -9 can stop anything dead in its tracks.  Once you see a valid block all current mining instructions should be halted and restarted with new latest block.

Offline bytemaster

Re: Mining on Linux instructions?
« Reply #40 on: November 06, 2013, 01:11:07 am »
I just found lots of 'generated block is stale' in my debug logs.  Is this normal?  I'm assuming this means the miner found a block but right before it found it someone else found a block.  I wouldn't expect to see this very often unless blocks were comin g really fast?  But I found like 10 of these errors.

This is the nature of the proof of work.   Unlike all existing POW systems, this system has momentum to it and so it cannot 'turn on a dime' when a new block comes in.   Some of these things may be because of fast proof-of-concept implementation, but some of it is the nature of the hash search taking several seconds to get any result.   
For the latest updates checkout my blog: http://bytemaster.bitshares.org
Anything said on these forums does not constitute an intent to create a legal obligation or contract between myself and anyone else.   These are merely my opinions and I reserve the right to change them at any time.

Offline bsamjolake

  • Newbie
  • *
  • Posts: 14
    • View Profile
Re: Mining on Linux instructions?
« Reply #39 on: November 06, 2013, 12:54:26 am »
I just found lots of 'generated block is stale' in my debug logs.  Is this normal?  I'm assuming this means the miner found a block but right before it found it someone else found a block.  I wouldn't expect to see this very often unless blocks were comin g really fast?  But I found like 10 of these errors.

Offline bytemaster

Re: Mining on Linux instructions?
« Reply #38 on: November 05, 2013, 09:08:15 pm »
You are memory bandwidth constrained.   It doesn't matter how much memory you have if the CPU cannot move data to and from it fast enough.
For the latest updates checkout my blog: http://bytemaster.bitshares.org
Anything said on these forums does not constitute an intent to create a legal obligation or contract between myself and anyone else.   These are merely my opinions and I reserve the right to change them at any time.

Offline smith88

  • Jr. Member
  • **
  • Posts: 45
    • View Profile
Re: Mining on Linux instructions?
« Reply #37 on: November 05, 2013, 09:04:30 pm »
For fun I even tried a 16core/32GB Linux, and it's hash rate is only 7.96353365. It's actually worse than my 8core/16GB instance running just over 8 hashespermin. CPU's are maxed out and memory usage is about 30%; it's running and connected to the network. Found 0 blocks in the past 10 hours.

Something seems wrong with the linux miner or there is some crucial config setting missing. Some guy on bitcointalk.org said he was getting 60 HPM on a 16core/32GB instance. That's 6x my rate!

My conf contains just the RPC user and password and addnode=xxx with one of the IP addresses. Any tips?
Thats because you probabl did the config incorrect. You have to launch multiple instances to utilize all 24 cores. So in reality you are using only about 3 of your 24 cores which is why you haven't found a block yet.

Wich I could help.  Last night when this thing went live I launched a couple of servers, one a Windows using only one of the cores, and was getting 2-5.  But on my linux server which has 24 cores, I was only getting maybe 10.  And I also found no blocks.  I am wondering if people are already throwing botnets at this...?  Wouldn't be surprised.  I am fairly sure that also the code is more Intel happy?  My windows server is a Xeon.

I could see this happening.  In config I told to use 24.  May I ask how to launch multiple instances?  I am assuming setup multiple directories with daemon running from each...?  Thanks. 

*By the way in TOP I could see all cores busy with all cpu usage coming from ProtoSharesd...
PTS:  Pv5uGw6APp4UweLE8J1piqWM2i5xFGUHij
BTC:  1B2vsiqJhh5W1mrfJNcC4Zu2kWG3mWzJft

Offline super3

  • Sr. Member
  • ****
  • Posts: 260
    • View Profile
Re: Mining on Linux instructions?
« Reply #36 on: November 05, 2013, 07:41:18 pm »
For fun I even tried a 16core/32GB Linux, and it's hash rate is only 7.96353365. It's actually worse than my 8core/16GB instance running just over 8 hashespermin. CPU's are maxed out and memory usage is about 30%; it's running and connected to the network. Found 0 blocks in the past 10 hours.

Something seems wrong with the linux miner or there is some crucial config setting missing. Some guy on bitcointalk.org said he was getting 60 HPM on a 16core/32GB instance. That's 6x my rate!

My conf contains just the RPC user and password and addnode=xxx with one of the IP addresses. Any tips?
Thats because you probabl did the config incorrect. You have to launch multiple instances to utilize all 24 cores. So in reality you are using only about 3 of your 24 cores which is why you haven't found a block yet.

Wich I could help.  Last night when this thing went live I launched a couple of servers, one a Windows using only one of the cores, and was getting 2-5.  But on my linux server which has 24 cores, I was only getting maybe 10.  And I also found no blocks.  I am wondering if people are already throwing botnets at this...?  Wouldn't be surprised.  I am fairly sure that also the code is more Intel happy?  My windows server is a Xeon.

Offline smith88

  • Jr. Member
  • **
  • Posts: 45
    • View Profile
Re: Mining on Linux instructions?
« Reply #35 on: November 05, 2013, 06:23:38 pm »
For fun I even tried a 16core/32GB Linux, and it's hash rate is only 7.96353365. It's actually worse than my 8core/16GB instance running just over 8 hashespermin. CPU's are maxed out and memory usage is about 30%; it's running and connected to the network. Found 0 blocks in the past 10 hours.

Something seems wrong with the linux miner or there is some crucial config setting missing. Some guy on bitcointalk.org said he was getting 60 HPM on a 16core/32GB instance. That's 6x my rate!

My conf contains just the RPC user and password and addnode=xxx with one of the IP addresses. Any tips?

Wich I could help.  Last night when this thing went live I launched a couple of servers, one a Windows using only one of the cores, and was getting 2-5.  But on my linux server which has 24 cores, I was only getting maybe 10.  And I also found no blocks.  I am wondering if people are already throwing botnets at this...?  Wouldn't be surprised.  I am fairly sure that also the code is more Intel happy?  My windows server is a Xeon.
PTS:  Pv5uGw6APp4UweLE8J1piqWM2i5xFGUHij
BTC:  1B2vsiqJhh5W1mrfJNcC4Zu2kWG3mWzJft

Offline rando

  • Full Member
  • ***
  • Posts: 100
    • View Profile
Re: Mining on Linux instructions?
« Reply #34 on: November 05, 2013, 05:56:06 pm »
For fun I even tried a 16core/32GB Linux, and it's hash rate is only 7.96353365. It's actually worse than my 8core/16GB instance running just over 8 hashespermin. CPU's are maxed out and memory usage is about 30%; it's running and connected to the network. Found 0 blocks in the past 10 hours.

Something seems wrong with the linux miner or there is some crucial config setting missing. Some guy on bitcointalk.org said he was getting 60 HPM on a 16core/32GB instance. That's 6x my rate!

My conf contains just the RPC user and password and addnode=xxx with one of the IP addresses. Any tips?

Offline yago

  • Full Member
  • ***
  • Posts: 188
    • View Profile
Re: Mining on Linux instructions?
« Reply #33 on: November 05, 2013, 12:24:53 pm »
Is anyone running a Linux miner consistently reporting a HPM that isn't 0 or is over 10?

Currently mining on x86-64 Linux  and got about 4 HPM. No luck yet  :-\
http://bitsharestalk.org/donate.html  <---- Donate to the BitShares Forum ----> PforumPLfVQXTi4QpQqKwoChXHkoHcxGuA

Offline bsamjolake

  • Newbie
  • *
  • Posts: 14
    • View Profile
Re: Mining on Linux instructions?
« Reply #32 on: November 05, 2013, 12:00:37 pm »
And now the windows box has found one while the Linux miners which have waaaay more power still have nothing and have been running more than 3x as long.  I'm getting pretty convinced at this point that something actually is causing them to be mining extremely slow.  Even though they're all maxing all CPUs.

Is anyone running a Linux miner consistently reporting a HPM that isn't 0 or is over 10?

Offline bsamjolake

  • Newbie
  • *
  • Posts: 14
    • View Profile
Re: Mining on Linux instructions?
« Reply #31 on: November 05, 2013, 11:19:33 am »
all my cores are maxed out, but is there any way to verify it's mining correctly?

hashespermin in getmininginfo shows 0 and the hashespersec command also shows 0

And is there any way to find out how many blocks you've found other than checking your balance?

./protosharesd getmininginfo
{
    "blocks" : 38,
    "currentblocksize" : 1000,
    "currentblocktx" : 0,
    "difficulty" : 0.00000095,
    "errors" : "",
    "generate" : true,
    "genproclimit" : -1,
    "hashespermin" : 0,
    "pooledtx" : 0,
    "testnet" : false
}

./protosharesd gethashespersec
0

P.S. I love the idea of ProtoShares and BitShares

Yes, this should work, but takes 4 or 5 minutes before it reports figures.

find a block yet? i've been running (from hashespermin) sum of 14.5 or so. no block yet. hah keep digging. :)

Nope.  It'd be real nice to have the hashrate in getmininginfo fixed for Linux.  I'm running both a Linux and win32 miners.  The Linux machine is way way more powerful than the windows one yet the Windows miner shows about 12 HPM whereas the Linux one is usually 0, but sometimes shows 1.xx or 0.xx.  Just making me worried theres a bug and it's not actually mining with the full power of the machine.