rasputin
Newbie
Offline
Activity: 14
Merit: 0
|
|
June 22, 2011, 02:27:53 PM |
|
NL was back up a few hours ago (didn't remember to post on this forum from work). Problem was found and fixed. EU (DE) and US West (#2-sort of) are being readies for being added as servers this evening.
After these new servers come up, I will not be adding any additional servers. Leasing new servers is not sustainable and adds a lot of overhead to the My Account/API pages. After the new US West and DE servers come up, I'm going to begin working on getting setting up a custom server solution with a single high power server, split into multiple sub-server VMs and a load balance VM to balance the connections between them. I have been putting this off due to the high initial cost, but I can almost afford to get the hardware and begin testing.
Once I have this new setup ready and colo'd at a data center, eventually the pool will shift into "two" servers, US and EU. They will sync up rounds just like they do now. Each server will actually be equivalent to 4-6 of the current servers in capacity, and will be able to scale significantly better than the current setup. This means a single server going down will get caught by the internal load balancer and the work will be redirected to a different VM.
If we hit major performance issues before this new solution is ready, I will be capping registrations until it is ready. We should be able to handle roughly 3.5 TH/sec before that is needed.
Interesting stuff. Do you feel like sharing some info on server specs and load factors? I'd love to hear what the main bottlenecks are for a btc pool. Would the DE server happen to be a hetzner dedicated server, btw? Great value for money, might be worth considering if that's not what you're thinking of.
|
|
|
|
eleuthria (OP)
Legendary
Offline
Activity: 1750
Merit: 1007
|
|
June 22, 2011, 02:35:58 PM |
|
Interesting stuff. Do you feel like sharing some info on server specs and load factors? I'd love to hear what the main bottlenecks are for a btc pool. Would the DE server happen to be a hetzner dedicated server, btw? Great value for money, might be worth considering if that's not what you're thinking of. Yes, the DE server is with Hetzner. The main bottlenecks for the pool are actually the TCP stack overhead in dealing with pushpool<->bitcoind traffic, as well as bitcoind which has some performance issues when getworks/second start to get very high. New server will be running Dual Quad Xeon's, 16 or 32 GB of RAM, split into a 7 server VM system. 5x pool servers, 1 MySQL server, and one front-end load balance server to manage the internal communications.
|
RIP BTC Guild, April 2011 - June 2015
|
|
|
mike85123
|
|
June 22, 2011, 02:56:43 PM |
|
Sorry for PMing this to you eleuthria and posting here (just got my noob training wheels off and finally able to post elsewhere), but I wanted to see what the rest of the users thought of my ideas so you could gauge the demand for them (I think they are good, but others may not see the benefit). Couple of feature requests/ideas: - Definitely need a way to sort the workers alphabetically (doesn't have to be selectable, just always sorted). Now it is sorted by when you created the workers. I have mine labeled 1x1, 1x2, 1x3, 2x1, 2x2, etc where the first number is the PC number and the 2nd is the GPU number within that PC. So if I add another GPU to a pc, they arent grouped together which makes it a little difficult to see what is going on.
- Ability to set workers between active and inactive just for showing up in the stats and API. Example, I had a PC stop working and had to order a new PSU and it would be nice to go in and set the workers for that PC inactive until I get it back up and running. Could be as simple as a check box in the manage workers that changes from 0 to 1 and then setting the stats and API to only show workers with a status of 1...
- If you implement the active/inactive workers, it would be nice to have a way to permanently delete workers. before i came up with my numbering/naming scheme I created a few miners that I would love to just wipe out.
- why not come up with a forum for just BTC Guild members instead of relying on this one thread. I think it could be more productive, although i suspect you haven't done it because of the time required to run it
Let me and everyone else know if these feature could be useful to you. Maybe if enough people want them, we can ask eleuthria to see about implementing them. Thanks for everything. love the pool.
|
|
|
|
rasputin
Newbie
Offline
Activity: 14
Merit: 0
|
|
June 22, 2011, 02:58:56 PM |
|
Interesting stuff. Do you feel like sharing some info on server specs and load factors? I'd love to hear what the main bottlenecks are for a btc pool. The main bottlenecks for the pool are actually the TCP stack overhead in dealing with pushpool<->bitcoind traffic, as well as bitcoind which has some performance issues when getworks/second start to get very high. New server will be running Dual Quad Xeon's, 16 or 32 GB of RAM, split into a 7 server VM system. 5x pool servers, 1 MySQL server, and one front-end load balance server to manage the internal communications. Awesome, thanks.
|
|
|
|
BitMinerN8
|
|
June 22, 2011, 04:32:46 PM |
|
Sorry for PMing this to you eleuthria and posting here (just got my noob training wheels off and finally able to post elsewhere), but I wanted to see what the rest of the users thought of my ideas so you could gauge the demand for them (I think they are good, but others may not see the benefit). Couple of feature requests/ideas: - Definitely need a way to sort the workers alphabetically (doesn't have to be selectable, just always sorted). Now it is sorted by when you created the workers. I have mine labeled 1x1, 1x2, 1x3, 2x1, 2x2, etc where the first number is the PC number and the 2nd is the GPU number within that PC. So if I add another GPU to a pc, they arent grouped together which makes it a little difficult to see what is going on.
- Ability to set workers between active and inactive just for showing up in the stats and API. Example, I had a PC stop working and had to order a new PSU and it would be nice to go in and set the workers for that PC inactive until I get it back up and running. Could be as simple as a check box in the manage workers that changes from 0 to 1 and then setting the stats and API to only show workers with a status of 1...
- If you implement the active/inactive workers, it would be nice to have a way to permanently delete workers. before i came up with my numbering/naming scheme I created a few miners that I would love to just wipe out.
- why not come up with a forum for just BTC Guild members instead of relying on this one thread. I think it could be more productive, although i suspect you haven't done it because of the time required to run it
Let me and everyone else know if these feature could be useful to you. Maybe if enough people want them, we can ask eleuthria to see about implementing them. Thanks for everything. love the pool. Yes, these are all great ideas. @Eleuthria, when you have time it would be nice if you could check into getting the hide/unhide feature working again under Manage Workers. I have clicked it and unclicked it on a miner hoping to reset it. Either way it still show up on the My Account page and API queries which my monitoring app keeps kicking out "miner down" alerts... Thanks.
|
|
|
|
eleuthria (OP)
Legendary
Offline
Activity: 1750
Merit: 1007
|
|
June 22, 2011, 04:59:45 PM |
|
US West is getting extremely lagged right now due to another set of back to back (3x blocks in 5 minutes) calculations which have a very hard time running with the pool being run on the same MySQL DB. Will be working on getting the pool split from the web server this afternoon.
|
RIP BTC Guild, April 2011 - June 2015
|
|
|
IlbiStarz
|
|
June 22, 2011, 05:12:10 PM |
|
Failed to connect to US East BTCGuild.
My main rig can mine on BTCGuild but my second rig cant :O
|
|
|
|
mike85123
|
|
June 22, 2011, 05:15:39 PM |
|
Yes, these are all great ideas.
@Eleuthria, when you have time it would be nice if you could check into getting the hide/unhide feature working again under Manage Workers. I have clicked it and unclicked it on a miner hoping to reset it. Either way it still show up on the My Account page and API queries which my monitoring app keeps kicking out "miner down" alerts... Thanks.
I guess I misunderstood the hide feature to begin with. Seems as if that is how it was built from the beginning (although you make it sound like it doesn't work right now). I never hit hide, thinking it would be gone forever from the stats page. just created a dummy did the hide button and it said i could unhide it. Can't check to see if it worked, b/c the site seems to be down (just hit a few blocks so he said it would take a while for the calculations to go through). Now I would love to see if it were possible to wipe out a worker. Wondering if it can't be done because he may track which shares were done by which worker, thus unable to delete any workers that have done any work. Maybe the move is to rename them to z0001 and hide them at the bottom of the list. @BitMinerN8, what are you using as your monitoring app?
|
|
|
|
IlbiStarz
|
|
June 22, 2011, 05:18:23 PM |
|
Does mining using just btcguild.com work?
|
|
|
|
PcChip
|
|
June 22, 2011, 05:25:26 PM |
|
main page down?
also I was having issues with uswest a few hours ago, so I ran double miners on each gpu, each with one connected to east, and one to west.
Question because I don't know much about how server hardware and vm's work: Why would you want to run seven VM's on that machine, each running the same exact thing? Why not just run one main OS and one main instance of the program? Or why not one main OS and seven instances of the program you want (bitcoind?) ?
Or is the main issue the TCP stack overhead, and using seven VM's is a way to 'cheat' around it ?
Thanks for any answers you can provide, I'm sure I'm not the only one who's curious!
|
Legacy signature from 2011: All rates with Phoenix 1.50 / PhatK 5850 - 400 MH/s | 5850 - 355 MH/s | 5830 - 310 MH/s | GTX570 - 115 MH/s | 5770 - 210 MH/s | 5770 - 200 MH/s
|
|
|
marvinmartian
|
|
June 22, 2011, 05:30:22 PM |
|
Just had a hiccup on the main page too (account to be specific). But it's back now.
|
"... and the geeks shall inherit the earth."
|
|
|
ensign_lee
|
|
June 22, 2011, 05:59:01 PM |
|
Does someone have 300 gigahashes / sec ? That's how much US West just lost. O.o Or did we bring down the electricity grid of a small city that had a lot of miners or something lol.
|
|
|
|
Litt
|
|
June 22, 2011, 06:04:42 PM |
|
Sorry for PMing this to you eleuthria and posting here (just got my noob training wheels off and finally able to post elsewhere), but I wanted to see what the rest of the users thought of my ideas so you could gauge the demand for them (I think they are good, but others may not see the benefit). Couple of feature requests/ideas: - Definitely need a way to sort the workers alphabetically (doesn't have to be selectable, just always sorted). Now it is sorted by when you created the workers. I have mine labeled 1x1, 1x2, 1x3, 2x1, 2x2, etc where the first number is the PC number and the 2nd is the GPU number within that PC. So if I add another GPU to a pc, they arent grouped together which makes it a little difficult to see what is going on.
Let me and everyone else know if these feature could be useful to you. Maybe if enough people want them, we can ask eleuthria to see about implementing them. Thanks for everything. love the pool. I would love to see this implemented as I hate having unsorted workers and this problem remains for other majors pools as well. This feature would set apart BTC guild for me from the rest making even better than it already is.
|
|
|
|
Carnth
|
|
June 22, 2011, 06:17:42 PM |
|
Or is the main issue the TCP stack overhead, and using seven VM's is a way to 'cheat' around it ?
Yes, the more servers you have the more TCP stacks you have. Also, because eleuthria will be placing load balancers in front of each VM an individual VM can be brought down for service with out affecting the system that much as a whole. It's kind of like spreading your eggs in different baskets.
|
|
|
|
BitMinerN8
|
|
June 22, 2011, 06:27:08 PM |
|
Yes, these are all great ideas.
@Eleuthria, when you have time it would be nice if you could check into getting the hide/unhide feature working again under Manage Workers. I have clicked it and unclicked it on a miner hoping to reset it. Either way it still show up on the My Account page and API queries which my monitoring app keeps kicking out "miner down" alerts... Thanks.
I guess I misunderstood the hide feature to begin with. Seems as if that is how it was built from the beginning (although you make it sound like it doesn't work right now). I never hit hide, thinking it would be gone forever from the stats page. just created a dummy did the hide button and it said i could unhide it. Can't check to see if it worked, b/c the site seems to be down (just hit a few blocks so he said it would take a while for the calculations to go through). Now I would love to see if it were possible to wipe out a worker. Wondering if it can't be done because he may track which shares were done by which worker, thus unable to delete any workers that have done any work. Maybe the move is to rename them to z0001 and hide them at the bottom of the list. @BitMinerN8, what are you using as your monitoring app? I have been using BitWid for Android. ( https://market.android.com/details?id=bitwid.qed.net) It was working great for a little over a week, but then I believe Eleuthria added more code to the API to return server loads, well that has caused the app to fail. (no fault to Ele, more stats the better) Then with Mt.Gox being down, none of those stats are working. So it's pretty much a dead app at the moment, but for $1 it is still worth it once API's are fixed. I have emailed the author and he said he will be fixing it soon, he wants to wait for Mt. Gox to come back up. I also suggested that he add TH and B7 to the list of Exchanges. He seems pretty responsive and he fixed and added some of my suggestions already.
|
|
|
|
mike85123
|
|
June 22, 2011, 06:32:51 PM |
|
I have been using BitWid for Android. ( https://market.android.com/details?id=bitwid.qed.net) It was working great for a little over a week, but then I believe Eleuthria added more code to the API to return server loads, well that has caused the app to fail. (no fault to Ele, more stats the better) Then with Mt.Gox being down, none of those stats are working. So it's pretty much a dead app at the moment, but for $1 it is still worth it once API's are fixed. I have emailed the author and he said he will be fixing it soon, he wants to wait for Mt. Gox to come back up. I also suggested that he add TH and B7 to the list of Exchanges. He seems pretty responsive and he fixed and added some of my suggestions already. Anyone find a good app out for Windows that parses through APIs at a set interval and will either pop a balloon up or send an email based on certain criteria (eg. last share greater than 5 mins ago).
|
|
|
|
gentakin
Member
Offline
Activity: 98
Merit: 10
|
|
June 22, 2011, 07:02:12 PM |
|
The stats API is broken right now. (I was wondering why BTC Guild disappeared from bitcoinwatch.com. ) http://www.btcguild.com/pool_stats.phpOnly current round time seems to be working.
|
1HNjbHnpu7S3UUNMF6J9yWTD597LgtUCxb
|
|
|
Bitonetta
Member
Offline
Activity: 66
Merit: 10
|
|
June 22, 2011, 07:50:57 PM |
|
eleuthria, nothing about delete workers? =/
Yes, delete and as mentioned a reorder button. I created a bunch of workers when I first started with poclbm. One for each card. Well, my farm has grown kinda large and it's easier to group them by machine for me now. Plus I started using hashkill which only uses a single login anyways. And my worker page is full of 'old' workers. They are hidden but I would just as well rather they just be deleted and alphabetized.
|
|
|
|
no_alone
Member
Offline
Activity: 84
Merit: 10
|
|
June 22, 2011, 09:51:35 PM |
|
You client say 132662 blocks??? give it some time..
|
|
|
|
|
|