|
The patched bitcoind did not contain a special command that I added to the old bitcoind (listgenerations rather than listtransactions). This meant when US West solved a round, the pool couldn't match it to a block, so it just kept going with a new round, causing a backlog of a few blocks. The way the pools synchronize the rounds means that it simply shifted the order of the block finds around.
I'm preparing to consolidate some of our servers this weekend. EU (DE) is our strongest server by far, and was able to handle 1 TH/sec with the current bitcoind patch before it began having idle miner errors. I have ordered a second server with the same specs which will come online tonight or tomorrow. It will replace NL1 and NL2. DNS records will be swapped over a few hours in advance so the vast majority of miners should never notice the switch.
I'm trying to locate a carbon copy of the specs (or as close as I can get) in the US to do a similar consolidation.
|
|
|
|
I was in the middle of wondering wtf when el posted. Unfortunaly, I'm still getting loads of errors. Hope it's fixed soon.
You also need to explain about uswest? You mean the server is going away? Will people have to go and switch their miners to point to another new host again?
Tomorrow I will be modifying the DNS to point uswest away from that server. After a few hours (well past the TTL), the pool will be shut down on the uswest server, leaving only the website. East, Central, NL, NL2, and DE can all handle significantly more volume now before choking, so the removal of US West's pool server will not be an issue. |
|
|
|
|
US West, East, and Central have been restarted with the bitcoind patch. West is kicking in some DDoS protection right now from the miner reconnect flooding, so the website has been temporarily taken down to avoid any calculation/communication errors between the pools. West will be back up shortly along with the website.
|
|
|
|
|
Stress testing is complete for JoelKatz' bitcoind patch. The server was able to run around 1 TH/sec before tossing out miner idles. NL1 has been returned to its original location, while NL2 is remaining pointed at EU (DE) for the time being.
NL1 and NL2 have had the bitcoind patch pushed live so far. US East and Central will receive the patch this afternoon. US West pool will be closed off tomorrow night now that we have a patch that drastically increases each server's capacity. This will speed up the site in general, and also stop new pool calculations from slowing down people that used to connect to uswest.
If we're lucky, JoelKatz (now known as our Savior) might be able to locate the new bottleneck within bitcoind to push it even further.
|
|
|
|
|
1 TH/sec is where idles started coming back for my DE server, which is an i7-920 system. I believe the new bottleneck is the getwork processing.
|
|
|
|
|
800 GH/sec and counting on the BTC Guild stress test.
UPDATE: 900 and counting.
BTC Guild, where test runs are done on production servers!
|
|
|
|
Any chance of the "You generated x.yy BTC in the last 24 hours." stat being exposed through the API?
Will be added shortly. Currently stress testing EU (DE)'s patch to bitcoind by pointing NL1 at the server. So far the results are looking good. |
|
|
|
Not that it matters much with the huge difficulty, but what happened to the 'Blocks found' column? I know it doesn't really do anything and mostly useless but it was nice to know how many winners your workers were responsible for.  Blocks found hasn't been working properly since the split to multiple servers. The information is there, but spread among all of the servers currently. I'll probably get around to setting up the servers to synchronize block finds this weekend so that Blocks Found and Hall of Fame can return. UPDATE: A couple quick restarts on EU (DE) to push a patched bitcoind that uses threaded RPC calls and keep alive connections. Stress testing the new setup now. |
|
|
|
|
UK's load balance service crashed and failed to restart overnight. The new block allocation script assumes when a pool server returns 0 shares for a round, there was a sync error and it aborts calculating the block.
Taking UK offline, it's causing a lot of problems and part of it is due to using a software level load balancer for a job that truly needs either a hardware or custom kernel/OS level balancer.
UK's addresses are now pointing at NL2 for the time being. Going to work on getting a similar setup to the server I ordered (should be delivered by next Monday), but as a dedicated host lease in the EU until the US server is setup, tested, and running live.
UPDATE: Banned an entire range of IPs from Brazil. It has been hitting NL1/NL2 hard the last few days, and now the entire range has been banned to speed up the auto ban process.
|
|
|
|
|
API/Website were having connection problems, which would be why your charts freaked out. UK cluster had an error in their sync code when they hit US East. It ended up flooding the server with MySQL connections which made it so the other servers could not get a response, making API/My Account not work for about 20 minutes.
|
|
|
|
Starting a temporary thread for BTC Guild until the original thread can get restored. Keep questions/comments in here until the main thread is restored, thanks. Brief Summary: 0% Fee Pool Long Polling Proportional 3 TH/sec Instant payouts even on invalids for 2.5% donators Website: http://www.btcguild.comPool Servers uswest.btcguild.com - California uscentral.btcguild.com - Texas useast.btcguild.com - Florida nl.btcguild.com - Netherlands #1 nl2.btcguild.com - Netherlands #2 de.btcguild.com - Germany uk.btcguild.com - United Kingdom |
|
|
|
btcguild having issues with proper updates on rounds again?
Rounds have been running fine all morning without any need for intervention. What are you referring to? |
|
|
|
|
The ads only appear for < 0.5% donators. Adblock or not, they will not appear unless your donation is at the lowest levels. The site is not "plastered with ads". It has one wide but short google adsense banner ad.
Other ways to make money? I'd be all for it! This was just a way to get something out of the continued growth. The pool's donations per block have been slipping consistently, and these ads on 0.0-0.4% donators are a very good way to normalize the revenue coming from people who don't give back to a pool that is currently demanding many dedicated servers.
|
|
|
|
|
Just sent the first of possibly many donations. Great efforts so far, and I'll be testing these on one of my servers tonight.
|
|
|
|
|
EU (UK) having a few issues after trying to point regular 'btcguild.com' to it (~600 GH worth of people who haven't picked a specific server). Working on diagnosing the problem, btcguild.com pointed back to DE for now.
|
|
|
|
|
Checking in after getting linked back here this morning. Is the patch now working in a high volume production environment? If so I'll have to throw some BTC at JoelKatz as well.
|
|
|
|
UK servers aren't showing share stats on the front end. We are getting credit & total shares show on the block stats page. I don't mean to be that guy but this has been the case with EVERY new server added recently... maybe its time for a check list Fixed the problem. It actually wasn't a problem on NL2 or DE. I added the stats -before- I put the server online this time. Unfortunately, I had a typo on the IP address (left off the last digit) used to request the server's worker summary for a user on the My Account page. im mining at eu.btcguild.com:8332/ and on my mining rig my Mhash/sec is 1670. but on btcguild it dont show my total Mhash/s or Estimated Rewards.
is it counting my shares or should i change pool?
Shares were still being counted, it was just the worker summary script not properly polling the UK server cluster. |
|
|
|
|
UK server cluster is now online. Current setup is two pool nodes. If all goes well, I will add a third node and take down NL2.btcguild.com (pointing that old address to the new server).
UPDATE: UK servers had a quick restart (~5 seconds) to fix their addnode settings so they will always connect to the BTC Guild pool nodes.
|
|
|
|
|
US West #2 is being pushed back in favor of a new EU (UK) server. This new server is actually a small cluster, one IP with multiple servers behind it. It should be up in the next 8 hours. If the load balanced cluster works, I will prepare to consolidate the NL pools into the cluster, eventually leaving us with two EU servers (EU DE and EU UK).
|
|
|
|
Long load time is due to the API having to grab live user stats from all of the servers. It will remain a long load time until the new server hardware is setup and ready to replace all the current US servers. |
|
|
|
|
Show Posts
