Flexible mining proxy

[exfuga]

Love, love, love the idea!

One thing, on the main admin page I get this error, and it only displays half of the table.

"This page contains the following errors:

error on line 37 at column 13: Extra content at the end of the document
Below is a rendering of the page up to the first error."

I am on PHP 5.3.2

[Papipapito]

Hello

Have a problem with a miner Ufasoft, with diablo and poclbm work fine

The miner connecting to proxy and start a work, but  in 5 minutes generate a timeout and miner stop

the log of apache:


 - - [08/Jun/2011:18:06:51 +0200] "POST / HTTP/1.1" 401 484 "-" "Ufasoft bitcoin-miner/0.10 CPU/Intel(R) Pentium(R) 4 CPU 3.00GHz/f43 PC=2 CC=1"
 - 2 [08/Jun/2011:18:06:51 +0200] "POST / HTTP/1.1" 400 392 "-" "Ufasoft bitcoin-miner/0.10 CPU/Intel(R) Pentium(R) 4 CPU 3.00GHz/f43 PC=2 CC=1"
 - 2 [08/Jun/2011:18:07:11 +0200] "POST / HTTP/1.1" 200 1015 "-" "Ufasoft bitcoin-miner/0.10 CPU/Intel(R) Pentium(R) 4 CPU 3.00GHz/f43 PC=2 CC=1"
 And now work and in 10 minutes, timeout and not appears nothing in logs

Any ideas?



Sorry for  my bad english

[tito13kfm]

So awesome!  Donation sent!

I look forward to future updates.  Would love graphs!  I don't even know what kind, but I love graphs!

[pwnyboy]

bump, seriously if this technical problem is solved and timing-out auth issue is fixed, I put a 5btc bounty on it.

edit: nvm I fixed it.

For the benefit of others, and because it's unclear whether a long polling problem still exists with the proxy code, would you care to share the nature of your problem and the solution?

[Azelphur]

Anyone else care to check the contents of their /var/lib/php5? after running this for a while I started getting no disk space errors, upon further investigation I found out that PHP had created 15.5 million session files in /var/lib/php5. Fun.

Should anyone else bump into this problem, you'll probably notice that nothing can delete the files (rm, ls, find etc all hang) after much fiddling, the solution is to use ls -U1, which will print the file list unsorted, and then pipe the output to rm, like so. ls -U1 /var/lib/php5 | xargs rm -f

Gonna take me 5 days of rm to clean up all the session files, fun fun

Besides this bug (which may just be some issue on my end) it seems great, I love how I can have pool redundancy in case one goes down.

[btcLeger]

Anyone else care to check the contents of their /var/lib/php5? after running this for a while I started getting no disk space errors, upon further investigation I found out that PHP had created 15.5 million session files in /var/lib/php5. Fun.

Should anyone else bump into this problem, you'll probably notice that nothing can delete the files (rm, ls, find etc all hang) after much fiddling, the solution is to use ls -U1, which will print the file list unsorted, and then pipe the output to rm, like so. ls -U1 /var/lib/php5 | xargs rm -f

Gonna take me 5 days of rm to clean up all the session files, fun fun

Besides this bug (which may just be some issue on my end) it seems great, I love how I can have pool redundancy in case one goes down.

cannot confirm this. have the proxy running under xampp. Yes, there are loads of sessions generated in the tmp-folder (about 10 per minute per miner) but expired session are getting deleted by the system. Must be a misconfigured serverside variable. Its not uncommon, Ive seen this often on some webservers when millions of sessionfiles slow down webservices. But dont ask me for a solution. Thats work for those who configure servers... 

[btcLeger]

Another suggestion for improved usability: I find it a bit unhandy to add new workers and pools when already having lots of workers & pools in the database. Would be nice to have a worker-edit page to configure all existing pools at once without the need to edit them all individually. Same with the pools. when adding a new pool, would be nice to update all existing workers for the pool on one page instead of picking them individually. Thats 30+ mouseclicks, where 3 would be sufficient...

[jondecker76]

I just finished setting up the bitcoin-mining-proxy.  I have the mysql database working, and have the pools and workers set up.  Now I have a couple of questions!

1)  When I go to my <ip_of_proxy> it asks for username and password, but they don't work there like they do at <ip_of_proxy>/admin.  Is this normal?

2)  How do I point my miner to the proxy?  I couldn't find anything in the documentation??  Since its on the same host, I'm sure its going to be 127.0.0.1 - but what port??  If I use 8332, it will interfere with the locally running bitcoind, correct?

So far very impressive!  Help me to point my miners (running phoenix r100) to the proxy and I'll post back results!

thanks again

[Cdecker]

I just finished setting up the bitcoin-mining-proxy.  I have the mysql database working, and have the pools and workers set up.  Now I have a couple of questions!

1)  When I go to my <ip_of_proxy> it asks for username and password, but they don't work there like they do at <ip_of_proxy>/admin.  Is this normal?

2)  How do I point my miner to the proxy?  I couldn't find anything in the documentation??  Since its on the same host, I'm sure its going to be 127.0.0.1 - but what port??  If I use 8332, it will interfere with the locally running bitcoind, correct?

So far very impressive!  Help me to point my miners (running phoenix r100) to the proxy and I'll post back results!

thanks again

1) the /admin you use the user and pass from your config, while / is for the workers to contact, so there you'd use the miners username and password
2) Since you run the proxy on a webserver you'll probably have to point the workers to http://<ip>:80/

HTH,
cdecker

[jondecker76]

I thank you, one Decker to another

Worked great!  Seamless failover for me as deepbit is priority 1, but it is down so its working great with the next pool in line - excellent!
Hash rate is normal - so far so good

[jondecker76]

is priority 1 higher than priority 2, or is priority 2 higher than priority 1?

[jondecker76]

didn't work well for me at all......  Had 3 pools set up, and 3 workers set up (I have 3 5850's, so a worker for each)..

priorities...
Worker1:
1 bitcoin.cz
2 btcguild.com
3 deepbit.net

worker 2:
1 btcguild.com
2 deepbit.net
3 bitcoin.cz

worker 3
1 deepbit.net
2 bitcoin.cz
3 btcguild.com


All 3 of my cards were hashing for bitcoin.cz for some reason, even though all 3 pools were up...  So each card should have been going to a separate pool........
The other problem is, after a few minutes, phoenix would display that the work queue was empty and hashes would drop to 0 (on all 3 cards/workers) and hang there....

So, so far, not so good unfortunately
I now have each card manually hashing each of the 3 pools listed above, and they all work fine..... Just not through the proxy.
Wil experiment some more later

[kjj]

is priority 1 higher than priority 2, or is priority 2 higher than priority 1?

2 will be used before 1.

Better idea is to space them out by 10s, or even 20s, so that you can squeeze pools in between without having to renumber them all.

[jgarzik]

See also https://github.com/kylegibson/poclbm for a poclbm fork that offers fallback support.  A bit easier than a proxy.

[kjj]

didn't work well for me at all......  Had 3 pools set up, and 3 workers set up (I have 3 5850's, so a worker for each)..

priorities...
Worker1:
1 bitcoin.cz
2 btcguild.com
3 deepbit.net

worker 2:
1 btcguild.com
2 deepbit.net
3 bitcoin.cz

worker 3
1 deepbit.net
2 bitcoin.cz
3 btcguild.com


All 3 of my cards were hashing for bitcoin.cz for some reason, even though all 3 pools were up...  So each card should have been going to a separate pool........
The other problem is, after a few minutes, phoenix would display that the work queue was empty and hashes would drop to 0 (on all 3 cards/workers) and hang there....

So, so far, not so good unfortunately
I now have each card manually hashing each of the 3 pools listed above, and they all work fine..... Just not through the proxy.
Wil experiment some more later

Check your URLs, usernames, and passwords.

Every time that I've run into a problem like this, it was because one or more pools was down, or because I had botched my configuration.

[jondecker76]

very strange..  it works fine if there is only one pool, but when I add multiples, it simply doesn't work.  I have also noticed that my stale/rejected shares went up extremely high...  In deepbit, for the last month I have been under 1% stales.  Using the proxy, I'm at almost 20% rejected shares

I was really hoping that this would work

[pwnyboy]

very strange..  it works fine if there is only one pool, but when I add multiples, it simply doesn't work.  I have also noticed that my stale/rejected shares went up extremely high...  In deepbit, for the last month I have been under 1% stales.  Using the proxy, I'm at almost 20% rejected shares

I was really hoping that this would work

The stale shares you're referring to are because of the aforementioned long-polling issues.  Nobody is sure what makes it work, or not work, through the current codebase.  The best suggestion I can give you is to run something like PHProxy at http://url.of.your.minining.proxy/phproxy/ and edit the code to send out this URL to your miners, rather than trying to use the in-built broken proxying code.  At least I think that'll work, from what I recall of the code and the way that LP works.  If you go this route, report back and let us know how it works out.

[cdhowie]

Anyone else care to check the contents of their /var/lib/php5? after running this for a while I started getting no disk space errors, upon further investigation I found out that PHP had created 15.5 million session files in /var/lib/php5. Fun.

Should anyone else bump into this problem, you'll probably notice that nothing can delete the files (rm, ls, find etc all hang) after much fiddling, the solution is to use ls -U1, which will print the file list unsorted, and then pipe the output to rm, like so. ls -U1 /var/lib/php5 | xargs rm -f

Gonna take me 5 days of rm to clean up all the session files, fun fun

Besides this bug (which may just be some issue on my end) it seems great, I love how I can have pool redundancy in case one goes down.

Thanks for the report.  This was indeed a bug, and has been fixed in master.

[kjj]

If you aren't using /var/lib/php5 for your session files, do "php -i | grep session.save_path" on the command line, or make a file with:

Code:

<?php
phpinfo();
?>

In your document tree, and load it in your browser, then look for session.save_path.

An alternative removal method is:

Code:

find /tmp/ -name sess_\* -exec rm {} \;

[cdhowie]

very strange..  it works fine if there is only one pool, but when I add multiples, it simply doesn't work.  I have also noticed that my stale/rejected shares went up extremely high...  In deepbit, for the last month I have been under 1% stales.  Using the proxy, I'm at almost 20% rejected shares

I was really hoping that this would work

That's weird.    The code path isn't any different if you have one or more pools configured...

If any of you are still having stale share issues, would you mind taking a capture of the network traffic?  On the proxy host, issue this as root:

Code:

tcpdump -w proxy.pcap -s 65535 'port 80 or portrange 8332-8337'

Then email me the proxy.pcap file after an hour or two.  Note that this file will contain all of the traffic on these ports and this means that I will have access to your proxy worker passwords as well as your pool worker passwords.  You may certainly encrypt the file to my GPG key if you are worried about the file being intercepted.  (Specifically, note that if you talk to a bitcoind from the web server, this will expose your RPC password!  So be especially careful in this case.  I'm not going to misuse this information, but... well, if I were you, I wouldn't trust me blindly either. )

If possible, capture traffic from the workers to the pools directly, without the proxy, and email me that as a separate file.  Comparing the two files will help me determine what's going on.

Even better, if you're using a pool supporting long-polling, have one miner going through the proxy and at the same time another one not using the proxy, and mail me both files.  Seeing how the long-polling requests return in both cases (they should be identical...) will help me diagnose this.

Thanks all for the feedback, donations, and support!  It's heartening to know that the software is fulfilling its purpose and that people find it helpful.