Topic: [CLOSED] BTC Guild - Pays TxFees+NMC, Stratum, VarDiff, Private Servers - page 305. (Read 903150 times)

This is true, I remember that several days ago I had a 50% more payout day from a serial of shifts that contained more than 25 blocks, now it's pay back time  

As I understand it, the issue was high CPU usage on bitcoind leading to invalid *shares*, not actually invalid *blocks*. Or was there another issue I missed? Here's slush's post from the slush pool thread:

It looks like it was a bug in Bitcoin software ... see message from slush pool "21.05.2013
Bug in bitcoind caused many invalid blocks generated in recent hours. Thanks to cooperation with bitcoin developers, a bugfix has been deployed. Pool is now back in normal operation."

You just burned electric power for nothing ...

This is also the problem of the Pool managers (generally young people ...) They roll rapidly new software build and forget the good old sentence (if it's not broken, don't fix it)

the old programmer.

24 Hour Earnings 0.09421285 from 1.58 estimate.

Really horrible at moment 

could it be something happened after everyone started accusing pool of doing 51%?

24 Hour Earnings 1.42611425 from 2.42 theoretical. There is just no way there isn't a problem.

I would say the pool is having seriously bad luck tonight, just saw a single digit of blocks from a shift ended in my PPLNS stats

Are you sure? How did you calculate this? I ask because I was just about to do the same thing and it's trickier than I initially thought. And also not so improbable.

Double posting just because I've been staring at this for so long that I'm actually feeling like I'm going through depression.  But thinking more clearly:

Just a few days ago, we had some VERY good luck.  A day or two before that, we had almost 20 hours straight of having shifts end between 10 and 60% above PPS.  Absolutely NOTHING has changed on the servers since that time aside from rolling the one server (which only has ~7% of the pool speed) back to 0.8.1 from 0.8.2.  So for it to be a software glitch causing the unusual bad luck we're having right now, you would have to expect that we should have had -even better- than that luck just a few days ago.

As I posted earlier, there were two "glitches" in this last week.  One was affecting many pools:  There has been a spam/dust "DoS" attack on the network that has been drastically slowing down bitcoind nodes.  BTC Guild does not currenlty use "empty" templates as a base when a new block is found, because the idea of making blocks with 0 transactions is just wrong.  However, as a result, this last week many of the pools were stuck for 4-6 seconds after each new block.  Many updates were being done to try to fix this, and recently sipa released a patch to drastically speed up creating new proper blocks.

The other glitch was one server [no others were found to be doing this] was crashing when submitting blocks.  I have not yet found out why it was doing this, but I rolled the bitcoind software back and it has not happened since.  That would not be a huge impact though, but it would contribute to the last few days of luck (5-10% maybe).



Very little has changed on the server side recently to cause any kind of software glitch (aside from the server that was updated to 0.8.2).  I'm actually starting to wonder if perhaps the problem lies outside the pool, possibly in the Avalon implementation of cgminer.  The luck has been trending negative since the time Batch 1 started showing up in large amounts.  I'm fairly certain BTC Guild has the highest % of Avalon users as a portion of the overall pool speed as well.

With the utmost respect for your expertise and knowledge, I have to ask..  Are you positive this is not a glitch in the system somewhere? The mathematical improbabilities of this happening are getting very large.

4 zeroes.. has that ever happened before, no amount of luck can compensate for this drought...

There is no crashing.  There's all zeroes in the last 3 shifts (currently) because we haven't found any blocks since those 3 shifts closed.

maybe you should be working on fixing that stratum server that crashes, since its all zeroes in blocks
 

I reckon regardless of what sort of long term luck graph you use, showing confidence intervals for the data is the thing that will ease most miner's mind's - if they can tell when bad luck was likely or unlikely they're less apt to worry about it.

I'm working on a more long term luck history graph which will hopefully ease the tension a little.  The luck has been trending negative for a while now, but in the big picture it is still quite close to neutral.  Only issue I'm running into right now is due to the way PPLNS handles share recording, it doesn't directly link to a difficulty like it did with PPS.  It's making it a little more difficult to create an accurate chart.

Well just my luck, just as I had my new miners the luck in the last 48 hours is awful haha. 

After a lot of log browsing, I have found a possible cause for the below average block solves in the last ~36 hours.  One Stratum server was moved to 0.8.2 as an attempt to combat the huge bitcoind latencies that have been plagueing pools lately due to a large amount of spam/dust transactions being generated.  For some reason, this server has occasionally been crashing when submitting a block to bitcoind.  It was not easy to identify when the crash hit due to the fact that the server was still submitting blocks (and solving them).  I've downgraded the server back to 0.8.1 now that we have a patch from sipa that dramatically speeds up the bitcoind process of creating a new block of transactions.

This was not the sole reason for the last ~36 hours, but it is likely a contributing factor.

PPLNS is having a terrible streak.

24 Hour Earnings 1.56 on 54GH  Theoretically 100% should be 2.42, which it was at before.

What on earth is going on, this is costing me big bucks