It was down for maybe...3 seconds?  I'd recommend holding your breath for 5 seconds and hitting refresh before running to the forum.

If you're getting idles frequently enough to make your 24h rewards less than projected (ie 0% luck), even during a period of positive luck, the problem is on your end.  Try moving to US West.  There is absolutely no way idles are eating up 30% of your mining time due to a problem on our end.  It would be hitting everybody equally, not just you.

Sounds like a routing issue between you and the server.  While I couldn't monitor them in the last 18 hours on my own miners, the last 3 hours have had 0 idles on my workers, on both US West and US Central.

Unless by "a lot" you mean one every couple of HOURS, in which case there's nothing that can be done.  Idles are caused by the flood of getwork/traffic requests that hit the servers after a long poll.  When they do hit, they tend to last less than 10 seconds.  So far they haven't shown up on my miners outside of when my personal connection was starting to fail.

Have you tried changing from west to central, or vice versa?

Everything is back to normal.  The error on the website was a known issue (which is why I was restarting it every 24 hours as a temporary fix), and everything is caught up.  Luck has continued to swing in our favor the last few days.  Funny how random numbers seem to be...random?

Sorry for the big website glitches this morning (maybe even last night).  My internet went down at home about 1 PM yesterday, but I wasn't home until 10 PM to find out it was actually my cable modem that went out.  Luckily I'm on a business line so they had somebody here by 8 AM to get it fixed.  I got the server back up and running around 8:15 AM thanks to my phone being able to SSH into the server and fix the problem.

I'm looking into why it's showing estimated shares negative right now.  I JUST got my internet back, but as far as I'm aware the pools themselves should've been running just fine, only the website frontend was having issues to my knowledge.

Don't worry, I'm working some magic on the blockchain.  You guys are going to like the next hour of reveals.  I'll give you a hint:  BTC Guild has done the last 4 blocks on the network in a row.

24H luck is a rolling window based on the SOLVED BLOCK END TIMES.  IE:  When a 6.8m share block gets knocked off the 24h list, the luck will shoot up because the average shares per block in the last 24 hours has just decreased significantly.

Payouts fixed, I've been having to restart the webserver quickly about once a day until the weekend when I can setup a new VM for the webserver, bitcoind didn't launch when it rebooted.

It looks like US Central had a few minutes of connectivity issues for some people.  Nothing hardware/software related, and seems to have already ended.

The servers are running as usual.  Its pushpool, bitcoind, and MySQL is sync'ing up without issue, I've had the most relaxing 2 weeks in a long time actually, other than our luck taking a dive which means all the conspiracy theorists come out of the woodwork to "question" things.  Like I said in my last post, everybody "questions" the bad, and ignores the good.  Anybody making the claim that our -14% this difficulty is unusual and not raising the same questions about DeepBit's positive 13.6% during this difficulty is a hypocrite.  If you want to believe that pool speed will eliminate the chance of variance "that high", then it's more than twice as unlikely for DeepBit to be pushing that luck as it is for us.

I just want to make sure we're all clear here.  When the luck bounces back and we're at +14% like DeepBit right now, and DeepBit is at -14% like we are right now, you guys better come in here shouting:

"I'm questioning if he's giving us free blocks to make us feel safe."  "I'm not accusing..."

While making the exact opposite accusations in DeepBit's thread.  Otherwise you're just hypocrites.

If you think -7.5% and our current difficulty luck is suspicious, I'd recommend you stay quiet if anybody ever asks your opinion on probability and statistics.  Everybody is quick to jump to conclusions when luck is negative.

I don't see anybody saying "I think he's adding fake blocks to make us happy" when Deepbit has had 12% positive luck during this difficulty while being MORE THAN TWICE our size, considering according to your post you think that having 2.0-2.4 Th/s should make such a huge luck (positive or negative)  should be "extremely unlikely.", which would infer having 5.1 TH/s should make it even more "extremely unlikely."

Our luck dived after the DDoS for two reasons, neither of which were related to DDoS directly.

There was significant server scaling issues in the end of June/early July.  Pushpools were crashing in the background and were sending getworks that had already been processed by somebody else before the crash.  On top of that the DDoS was causing us to be working on invalid block chains, which is why we had a lot more invalids than normal around blocks 1400-1600.  

On top of all that, new dev patches for bitcoind to help the servers scale were causing other problems.

Since the DDoS ended the stability problems have vanished, now planned downtime only takes a server off for a minute or two, and we haven't had any downtime related to an unexpected crash in almost 2 weeks (since US West was "finalized").

If we are in the green, it means our pool is getting lucky, and we're solving the blocks faster than normal.  This means more money in a 24 hour time frame than average.  The opposite case if we are in the red.

Fixed an error in the graphs which caused pool/user speeds to be a little erratic when a long block ended shortly after a new day began.  The old system averaged speeds using the following:

Shares in completed blocks during that day, divided into a full 24 hour day.  Since it only looked at completed blocks, if a long block ended at 12:01 AM, the previous day was being calculated with a large chunk of time/shares missing, and the next day had that large chunk of shares included.

New graphs divided it by the number of seconds between the start of the first block and end of the last block in that 24 hour period, providing a much more accurate estimate of pool speed/user speeds for the day.

Its how our luck has been over the past 24 hours (ie:  How many shares its taken us per block for the past 24 hours vs the current difficulty).

If average shares per block are higher than difficulty, bad luck ( red ).  If average shares per block are lower, good luck ( green ).  Nothing we can do about it, its just there to provide a comparison number to 24H rewards.

US West had a brief downtime.  I was hoping to do it without the pools going off at all, but it ended up being about 3 minutes in total.

Doing some prep work for this weekend so I can put US Central's load balancer setup onto US West.  The hope is that I can lay all the ground work down offline during the week, and this weekend I can do the switch with only about 1 minute of the US West pool being down.

Stay away from us, we're on a roll!

Added 24 hour luck back to the top stat bar, just so people have a reference point for 24H rewards.  I really hate seeing that number though.

Our speed gains are mostly related to people returning from namecoins.  Lots of people in our chat went to NMC for a few days to exploit that market for the brief window before NMC difficulty quadrupled.

All we can hope for is more pool speed to even our our variance, or a lower difficulty.  Other than that, its just praying to get a few days of good luck again to offset the last two days.

Looking at the block details, I'm seeing nothing wrong with the block you identified.  The number of users credited with shares that round is within the expected range of users for a round of that length.  In case of a sync error, assuming the failsafe didn't catch it, you'd see roughly half as many people getting rewards, and getting twice their expected average (ie: One of the two servers didn't sync up).