Topic: [~1000 GH/sec] BTC Guild - 0% Fee Pool, LP, SSL, Full Precision, and More - page 10. (Read 379078 times)

When working with databases it is always much easier and preferred to mark things disabled/inactive rather than outright deleting them because you end up with holes in the sequence/record of events happening in the system.

I just got another one 10 minutes ago at 11:58 Eastern

The idle email issue should be resolved as of about an hour ago.  As you mentioned, it was an error in the way in time zone handling.

I've been getting bogus idle miner notifications this morning.  Looks like the box that sends them is having a 9 hour disagreement with the database servers about the time (or time zone).

I'm planning on making some revisions to the website this week, mostly in an attempt to begin integrating the PPS pool with the primary website so that you share accounts/workers across the two payout methods.  Once the merge happens, you will still need to use pps.btcguild.com to connect to it, since its using different pool software at the moment.  Eventually updates to the pool software will be deployed and flag individual workers with PPS/Proportional flags so you can switch on the fly without changing your miner configurations.


While making the revisions, I will be trying to make everything more modular for easier editing, and part of that will include "Deleting" workers [permanently hide].  A lot of BTC Guild was designed in the fastest way that would make it work, rather than efficiency, since it was being built while the pool speed doubled every few days.


As a side note, the main pool will start integrating a few PoolServerJ nodes soon.  There was a brief test last night where approximately 10% of the pool was shifted to PSJ.  Performance was reasonable, though the load was a bit too much for our pushpool VPS nodes to handle.  However, I believe putting it onto a dedicated quad-core server with ample RAM, PSJ will be able to blow pushpool away.  The new server has been provisioned, I'm just waiting for a small long poll bug to be resolved before it gets deployed.

I know this has been said before, but I'm going to say it again...

I absolutely love your pool and I love everything about it, EXCEPT that you can't remove workers completely. Even if you can't put the option into the DB, then at least give us the option to Delete it and just set a Flag in the DB and don't ever show us that worker again. Later you can go back and remove all those deleted workers, but it can be quite a hassle trying to modify workers and remembering who is who when you screw up as much as some of us do.

We've been getting "attacked" technically for the past day.  The servers are more resilient than before due to load balancing, but it has caused some connection/idle issues.  The accounts are being automatically banned as they crop up, it hasn't been too hard to spot the malicious users, but it isn't possible to filter them until after they've caused some performance issues.

Things have been improving steadily for the last few hours.


Additionally, we've opened up a Pay-per-Share version of BTC Guild, accessible at http://pps.btcguild.com .  This is REAL Pay-per-Share, not SMPPS/PPLNS, meaning there is no concept of backpay/limits on your rewards.  You get 'X' per share (based on difficulty, with a fee of 7% compared to Deepbit's 10% PPS), and it is available for withdrawal immediately without any type of confirmation timer [0.10 minimum payout].

The new pool server is running on PoolServerJ.  Since this pool software hasn't been tested as thoroughly as pushpool, BTC Guild's PPS pool is considered beta.  Share submissions and rewards are secure, but if the pool crashes it may not automatically recover like our pushpool servers do.  As such, it is recommended you use a miner with the ability to failover to another server when using BTC Guild (PPS Beta).

Some new servers added to the round robin, dedicated for load balancing rather than LB+Pushpoold on the same server.  Should prevent the problems we were having yesterday.


UPDATE:  Doing a full reboot of all the servers.  Trying to make the change live seems to have caused very uneven loads.

All of the servers are online and running smooth again.  You said they failed over to ArsBitcoin an hour ago, which is when the servers were going up/down due to the large network that was basically DDoS'ing our servers systematically as they failed over from one server to another.

I'm hoping by tomorrow morning the dedicated load balancers are up and running, at which point I'll also be turning on the autoban/filter for the pools to prevent the issues we had this evening.

All five of my machines now cannot connect to uswest.btcguild.com:8332, they all failed over to arsbitcoin starting about an hour ago or so

Just banned a large IP range, looks like someone with access to a couple thousand computers tried to point them at the guild, and they weren't using the round robin DNS.  So they'd point at one IP, overload it to the point of failure, then move to another as they failed.


Word of advice if you're going to steal electricity from your university/corporation:  Learn how to use miners properly.  I don't care how many machines you're leeching, you don't need to issue 20,000 getworks/second.

Well, once we get that far we might as well get the record at least. 

E,  thank you for your comments and explanations.  we/I appreciate it.

Is US Central having problems?

I've identified the problem people are having connecting.  The first two servers in the round robin list are receiving far greater load than the other 8.  If you're having issues with your connection try changing to any of the following IPs directly:

78.46.186.29
78.46.186.30
78.46.186.50
78.46.186.51
78.46.186.52
78.46.186.53
78.46.186.54


I'm provisioning a pair of two small servers to put at the top of the round robin list to alleviate the problem.  Hopefully they will be online in just a few hours.

Yeah, during the final hour of the block (I was asleep for most of it), the only thought going through my mind was:  "Please don't have us set the new record for most shares in a single block."

Jesus


Shares per Block (24H)      3,304,038   

2872  144489   2011-09-08 13:53:06   7:12:33   11743359

8 minutes between long polls is absolutely normal.  You were idle for under 2 minutes.

The second set of disconnects/reconnects in your log were me restarting the load balancers, I added an extra server to the list [the old 2nd server for SC Guild].  You'll notice the first disconnect is less than 30 seconds before the last reconnect. 

I also banned 17 IPs last night that were showing very high getwork requests and low/non-existent share submissions, which should help as well.  I'm still manually performing these bans until I'm confident my script wont ban legit users that have clients keeping a connection open with the pool for failover/load balancing in cgminer/poclbm.

Eeek. Last 4 blocks have followed doubling progession:

0:30:02, 1:11:02, 2:03:50, 4:01:16 (so far).

I sure hope this doesn't continue   

WTF is this?. I can connect to other pools just fine so don't tell me it's a problem on my end.

Edit: Fixed it myself, I had to flush the dns cache and reboot the modem a couple of times before it finaly updated. What kind of ass backwards server did you move the pool to :/