Release issued today

[Steve Sokolowski]

Today, we issued a new release. Here's a list of all the changes that were made to the system since the beginning of the weekend.

• We discovered a major flaw where we there were strong references to some miners left over for long periods of time, which meant that they did not undergo garbage collection for a while after they disconnected. There were a number of effects from this issue, which has been present for at least a year and represents at least 50 hours of investigation.
• Miners who had disconnected but weren't cleared from memory messed with the weights of the coins being mined, causing the system to believe that miners were still mining some coins. Since coins tend to decline in profitability over time and those miners would never be reassigned, I suspect that this caused miners to be paid less than they should have, but I can't be sure. Whatever the case, this issue could not have caused miners to be paid less than the reported profitability, because that's just an average of what everyone is earning. I think the end result will be a slight increase in reported profitability compared to that over the past year.
• Performance was dramatically improved because now these "ghost" miners aren't being included in calculations. For example, I was able to remove a routine that periodically corrected summary tables that would get out of sync.
• All of the lost shares to this point were caused by this error. The error caused weights of coins to be present that could be in error due to having no price, since the miners who were mining those coins had disconnected.
• The next category of issues involves website changes, which I decided to focus on because we want to eliminate support tickets that might actually be caused by display issues rather than the mining server issues that I've been investigating to little success.
• First, I added a check to disconnect users who connect with two miners with the same n= parameter. This is important because n= is used to track miners on the website, and the hashrate of two miners with the same "n=" would constantly swap back and forth on the "earnings" page.
• Second, we added checks to hard disconnect users with an incorrect username, rather than just allowing them to connect and correct the issue later. There are at least 20 users per second being disconnected due to this issue now.
• Third, I discovered and fixed an issue where the share count over two hours never included the shares submitted during the past five minutes, so the count of shares would appear to be lower and converge on the actual number as time went on. The most immediate impact from this fix will be that some may notice that the share count begins increasing five minutes earlier than it would otherwise.
• I also added a line to send out an update when the first share is submitted. That way, instead of the website reading "no valid shares" for the first up to five minutes, now it will read one share for that period, along with "calculating hashrate." This is important especially for users who are mining static coins that have infrequent blocks and therefore won't see updates on the share count very frequently.
• We also discovered an issue that could cause new coins that were just added to be unavailable for mining without a restart. Now, even under that circumstance, new coins will always be available for mining immediately upon addition.

Feel free to offer your comments!

[mickeekung]

Interesting release!!!!

[xichas]

Ok Steve thanks for the hard work!!! I´m sure it´s not easy spending your weekend doing this, and I thank you and your team for that. Hope this changes and bugs found will solve the lack of satability and profitability, that is happening untill now. In the end we wiil all have more profit.
Regards and respect

Xichas

[caosmen]

Looks like it solved the Problem.
right now everything runs smooth again.
but lost today still

[GregoryGHarding]

wow steve, this update contains alot of fixes, im excited to test these out over the coming days, thanks alot for your fantastic work and devotion

[mycide]

Looks like a good fix!

Im still unsuccesful tring to solo mine Einsteinium however.

[Jalapan]

Im mostly negative about stuff but this is great!

Good work and keep it going .. im impressed.

Also are you aware of the display bug of electricity costs and found blocks not showing all blocks a user has found?

[Steve Sokolowski]

Im mostly negative about stuff but this is great!

Good work and keep it going .. im impressed.

Also are you aware of the display bug of electricity costs and found blocks not showing all blocks a user has found?

I know that there are a small number of blocks per day, about 10, that come from daemons that take over a minute to reply. Those blocks are dropped because there isn't anything else we can do. Fortunately, the daemons are all low-priced coins.

The electricity costs issues are low priority now. They'll be dealt with by the new people if we decide to expand once the lawyer replies.

[tomos81]

First, I added a check to disconnect users who connect with two miners with the same n= parameter. This is important because n= is used to track miners on the website, and the hashrate of two miners with the same "n=" would constantly swap back and forth on the "earnings" page.

in nicehash - now when we are trying to hire higher hashing power, and nicehash creating more mining groups within the same order ,the 2nd and the next miner groups having the error - Disconnected - Authorization failed..

https://ibb.co/gvqgTk

[hashingpro]

So if we had 100 people on, and the system was reading it as 120 miners due to not disconnecting 20 people and still counting them....

Then wasn't the system taking the profits and dividing it by 120 and giving each miner 1/120 of a profit instead of 1/100 of a share?
So where were the rest of the 20 1/120 shares going if not to the miners that were on the system mining?

Were the expected payouts being figured using this invalid 1/120 number instead of the true 1/100 number it should have been and THATS why all the expected payout projections have been off for so long now?

Hopefully with the disconnected people no longer being counted to skew off the numbers the expected payouts are more in line with what we actually get so the expected payouts can again be more reliably used to judge performance of the pool again.