Smartcoin Linux mining administration. [MULTI-MACHINE SUPPORT NOW IN!]

[jondecker76]

Did poclbm change recently confusing smartcoin?  I just installed poclbm (git://github.com/m0mchil/poclbm) and smartcoin for the first time. The syntax smartcoin was using (--host, etc) was invalid.  It also does not appear to be able to scrape the output (I can see the miners running in screen, but smartcoin does not read it).   Additionally, it was (kind of) using invalid device ids.  Poclbm uses device 0 as the CPU.  Therefore the first miner ended up on the CPU.  

I was able to work around all of these except the scraping issue.  I was just wondering.

Dan

Can you post the smartcoin status screen output, along with the output of poclbm from the miner screen? Perhaps something has changed, I did test it and it worked just a couple of weeks ago

Also, to my knowlege poclbm doesn't always assume device 0 is cpu (unless something has changed in newer poclbm revisions). Here is the output of me running poclbm with no parameters:

Code:

jondecker76@miner1:~/poclbm$ ./poclbm.py                                        
No device specified or device not found, use -d to specify one of the following 
                                                                                
[0]     Cypress                                                                 
[1]     Cypress                                                                 
[2]     Cypress                                                                 
[3]     Intel(R) Pentium(R) 4 CPU 2.80GHz                                       

[elrock]

What is the default miner output format string?  I tried editing the format string and ended up erasing it entirely.  Weird characters appeared every time I pressed one of the arrow keys or the delete key to move the cursor.  Now it's totally blank and I can't see a thing.

[jondecker76]

What is the default miner output format string?  I tried editing the format string and ended up erasing it entirely.  Weird characters appeared every time I pressed one of the arrow keys or the delete key to move the cursor.  Now it's totally blank and I can't see a thing.

The default is:

Code:

[<#hashrate#> MHash/sec] [<#accepted#> Accepted] [<#rejected#> Rejected] [<#rejected_percent#>% Rejected]

[elrock]

Thanks -- I found it by searching through the thread.  I know you've already got tons on your plate, but it might be helpful to new users to include that info in the first post.

[Fletch]

Note, that if you run an unusually high loop delay, you may want to reduce the Failover Threshold and Lockup Threshold settings as well, as they are measured in loop iterations.

I'll do that. My loop delay is 10 and I've set the failover threshold to 3 and the lockup threshold to 10. Would that make sense?

Perhaps you can add a sentence under "Edit Settings" describing what these settings actually mean?

I will look into other methods of determining CPU load, but an average works really well here, as the script its self has some areas of really high usage, and possibly even no usage at all (during the loop delay for example)  Can you post some comparisons of /proc/loadavg against iostat?

It's difficult to compare load averages with straight up CPU utilization. They're not measuring the same thing. I would say load averages are more interesting though as they reflect how the CPU (or CPUs) is coping with the work load. It's quite possible to run at 100% CPU without overloading the machine. It's also possible to overload the machine without running the CPU at 100%. More info here: http://en.wikipedia.org/wiki/Load_(computing)

[jondecker76]

kennel -
Your settings look good to me.  If you get any false triggers of failover and/or lockup, you can always increase them 1 or 2 at a time until you find a good balance.

I'll experiment  with the /proc load average tonight (just got to decide whether to go with 1 minute or 5 minute average, though I'm leaning towards 1 minute) Additionally,  it would get rid of another dependency.

Regarding better descriptions of the settings, I'm probably going to leave it documentation in the future.  The settings system is very dynamic in that I just add an entry into the database and it automagically appears and just works - this makes development a little easier on me.  I may rethink it a bit once I can get out of beta, as I won't be adding entries so often then and it may be worth extending the system to be more verbose.

[EnzoMatrix]

kennel -
Your settings look good to me.  If you get any false triggers of failover and/or lockup, you can always increase them 1 or 2 at a time until you find a good balance.

I'll experiment  with the /proc load average tonight (just got to decide whether to go with 1 minute or 5 minute average, though I'm leaning towards 1 minute) Additionally,  it would get rid of another dependency.

Regarding better descriptions of the settings, I'm probably going to leave it documentation in the future.  The settings system is very dynamic in that I just add an entry into the database and it automagically appears and just works - this makes development a little easier on me.  I may rethink it a bit once I can get out of beta, as I won't be adding entries so often then and it may be worth extending the system to be more verbose.

If you are using the database for that then you could add a description field utilize that for basic info with out too much modification to your existing code ( just a thought ) keep up the great work love the code so far

--Enzo

[EnzoMatrix]

If you are going to utilize the load / cpu averages make sure that the thresholds are adjustable as my mining box also runs a minecraft server so my cpu runs at around 30% - 35% pretty constantly.

[Fletch]

I'll experiment  with the /proc load average tonight (just got to decide whether to go with 1 minute or 5 minute average, though I'm leaning towards 1 minute) Additionally,  it would get rid of another dependency.

+1 for going with the 1 minute average.

[jondecker76]

Update r512e now available:
- CPU load is now determined by /proc/loadavg 1-minute average
- Huge lockup detection optimization..  30%-50% faster status loops and lower CPU load as a result!
- New "<<<FAIL>>> instance status output. This tells you that the miner failed to load for some reason (you can then verify with screen -r miner)

[EnzoMatrix]

in r512e there appears to be an issue with the status display of the auto donate function and it keeps flashing an error on refresh about =python command not found the error is reporting line 355 of smartcoin_ops.sh as the culprit

--Enzo

[jondecker76]

Update r513e now available
-Fixed small typo in the generation of the autodonation fieldarray. ThanksEnzomatrix!

[EnzoMatrix]

Have you ever considered an optional ncurses interface that would allow for some fairly nice console displays? I realize that this would cause for another dependency however if it was optional then it would give the end use the option of taking advantage of that or not.

--Enzo

[Fletch]

Not sure if this is doable, but whenever there's a failover or profile change, Smartcoin shuts down all miners and start them up again. Would it be possible for Smartcoin to keep track of which miners are running and when there's a profile update or whatever, it would only stop and/or start the miners affected by the change?

For example, if I'm mining with 3 GPUs at bitcoins.lc and there's a failover to deepbit, Smartcoin will shut down the 3 bitcoins.lc miners and start up 6 new miner instances (3 for bitcoins.lc and 3 for deepbit). The restarting of the bitcoins.lc miners in this case is redundant. If connectivity with bitcoins.lc is restored, Smartcoin will kill all 6 miners and start up 3 new bitcoins.lc miners, when in fact, just killing the deepbit miners would suffice.

[jondecker76]

Have you ever considered an optional ncurses interface that would allow for some fairly nice console displays? I realize that this would cause for another dependency however if it was optional then it would give the end use the option of taking advantage of that or not.

--Enzo

I think its a good idea... Once all core functionality is in, I may look into this!

[jondecker76]

Not sure if this is doable, but whenever there's a failover or profile change, Smartcoin shuts down all miners and start them up again. Would it be possible for Smartcoin to keep track of which miners are running and when there's a profile update or whatever, it would only stop and/or start the miners affected by the change?

For example, if I'm mining with 3 GPUs at bitcoins.lc and there's a failover to deepbit, Smartcoin will shut down the 3 bitcoins.lc miners and start up 6 new miner instances (3 for bitcoins.lc and 3 for deepbit). The restarting of the bitcoins.lc miners in this case is redundant. If connectivity with bitcoins.lc is restored, Smartcoin will kill all 6 miners and start up 3 new bitcoins.lc miners, when in fact, just killing the deepbit miners would suffice.

I've already done some experiments, and I will be going this route sometime in the future ("hot reloading") - its something I have been thinking about for a while, just haven't got around to it yet as the current method works just fine, but eventually this would be a great optimization!

[jondecker76]

Just a note not to do an update for a while (I'm uploading a bunch of VERY experimental multi-machine changes that *could* leave your inoperable for just a little while - I'll post back here in a little bit once they are all in and tested, at which time an experimental update will be relatively safe)

[EnzoMatrix]

Hello,

  I am not sure if it is something I am going wrong but I am unable to trigger the failover

I have a profiles setup to mine on 3 different server on the same pool and I have tried simulating an outage using /etc/hosts file but it never fails over is there any specific data that might help in locating this.. ? it is a fresh smartcoin install

Smartcoin r495s

here is the failover order
1 was a deleted profile

2) BTCGuild All
3) BTCGuild US
4) BTCGuild USWest
5) BTCGuild USEast
6) BitClockers

Found the solution apparently I ended up with 2 menu options / database entries for failover threshold that was confusing the system as soon as I deleted one of them failover started working.

--Enzo

[jondecker76]

Update r544e now available:
- Many changes to different areas of the code to support multiple machines.  Though multiple machine support isn't fully included yet, the "backend" functions are now machine-aware This puts multi-machine support probably near 98% complete. Just a couple of key components remain....  These were amongst the biggest changes needed yet for multi-machine support, and aside from some very small odds-and-ends, the only thing left to implement is a "Configure Machines" config screen option


NOTE: There is currently one side-effect that I haven't fixed yet.  The miner screen session's (screen -r miner) status line is all messed up, though it doesn't affect anything.  I'm going to continue to experiment until i get the status line back, but its really not even needed and only a visual artifact if you are looking at the miner screen session for now.

[DBordello]

+1 for option to exclude profiles from fallover (or better yet, just the ones specified).  I have some "experimental" profiles that I would like it to stay away from.

Dan