Re: [5000 GH] Eligius: Decntrlzd, ASIC-rdy, 0Fee CPPSRB, 0reg, BTC, 877 # support
Greetings Eligius miners!
So, I was up yesterday morning at 5:30AM local time (Saturday!) because of one of my "pool is broken" alarms. Turns out some insanely huge botnet has decided to call Eligius home, and is adding bots at a rate of about 2-3 unique IPs per second. And of course, these bots are connecting on port 8337 (getwork/GBT).
So, of course, after building up for a while, the server got bogged down and tripped my alarm. I've since been trying to figure out a solution to the problem, and it is now just about 27 hours later (I had about a 90 minute nap somewhere in there).
Anyway, after many many attempts at different methods for fixing this, I ended up coming up with a solution which basically amounts to load balancing port 8337. I've also moved stratum (port 3334) to it's own independent process.
Now, eloipool wasn't really designed to do this, so, I don't know if there are any unforeseen issues with the way I've attempted this. One side effect comes to mind, and it involves variable difficulty calculation for miners who mine from multiple IPs. Most likely their vardiff number will be lower than it normally would be because of this if they use GBT/getwork because they wont be getting/submitting work to the same server on the backend. Not a huge problem. One which I will solve soon.
Edit: I also just realized that this setup affects vardiff for anyone with hashpower on both GBT/getwork and stratum.
So, conclusion, the pool is up, its handling the new traffic now, and all should be well for miners.
-wk
So, I was up yesterday morning at 5:30AM local time (Saturday!) because of one of my "pool is broken" alarms. Turns out some insanely huge botnet has decided to call Eligius home, and is adding bots at a rate of about 2-3 unique IPs per second. And of course, these bots are connecting on port 8337 (getwork/GBT).
So, of course, after building up for a while, the server got bogged down and tripped my alarm. I've since been trying to figure out a solution to the problem, and it is now just about 27 hours later (I had about a 90 minute nap somewhere in there).
Anyway, after many many attempts at different methods for fixing this, I ended up coming up with a solution which basically amounts to load balancing port 8337. I've also moved stratum (port 3334) to it's own independent process.
Now, eloipool wasn't really designed to do this, so, I don't know if there are any unforeseen issues with the way I've attempted this. One side effect comes to mind, and it involves variable difficulty calculation for miners who mine from multiple IPs. Most likely their vardiff number will be lower than it normally would be because of this if they use GBT/getwork because they wont be getting/submitting work to the same server on the backend. Not a huge problem. One which I will solve soon.
Edit: I also just realized that this setup affects vardiff for anyone with hashpower on both GBT/getwork and stratum.
So, conclusion, the pool is up, its handling the new traffic now, and all should be well for miners.
-wk
Have you considered what happens to variable difficulty when multiple miners are using the same IP and the same payment address. For example one public IP is used for three miners behind a router and firewall but they have varying hash rates (180, 90, 70000) requiring difficulty 1, 1, and 64, respectively? This appears to be a problem even when using the new addresses for getwork for the first two and stratum for the third - where the slow miners usually end up getting difficulty set far too high and the fast miner occasionally has difficulty set too low. The same appeared to happen more than one miner used the new stratum address.