Smartcoin Linux mining administration. [MULTI-MACHINE SUPPORT NOW IN!] - page 16.

jondecker76

full member

Activity: 238

Merit: 100

Quote from: tigereye on July 18, 2011, 11:56:04 AM

Quote from: jondecker76 on July 16, 2011, 08:37:30 AM

tigereye

I just committed a change regarding the /tmp directory, perhaps it will help. smartcoin is meant to be installed and run as user. Here is what I would do:

1) Under user, make sure you clear the data directory:

Code:

rm -r ~/.smartcoin

2) Clear out any temporary data:

Code:

sudo rm /tmp/smartcoin*

3) Follow the install stuff again. Remove the old smartcoin directory and start again.

4) During the detection of the AMD/ATI SDK detection, there is a bug on some systems which shows two paths on one line. Since Linuxcoin is 64 bit, make sure to leave only the 64 bit relevant path!

5) If after starting things appear not to run, detach from smartcoin (option 3 on the control screen), then run

Code:

screen -r miner

This will bring you to the individual miner instances where you can see directly the error output of phoenix

Jon,

I've reinstalled smartcoin from scratch and re-setup all of my workers.
I was seeing the same symptoms as before (a bunch of "grep" error messages flying across the screen every few seconds, and no miners would stay up)

Then I created a manual profile and selected it, and it works!
I went back to the automatic profile, and it showed the same symptoms.

It seems something's funky with the 'automatic' profile.

Hope this helps you zero in on things. If you need any debug info to help narrow it down, let me know.

Can you double check one thing for me? Go to Configure Miners->Edit and edit the miner you are using. Make sure to answer "y" (yes) when it askes if it is the default miner. (The automatic profile uses the miner flagged as default - it is possible that one isn't flagged as default, or was never flagged initially)

Also, I'm committing a couple of changes that silence grep errors, as there are times during a profile change where grep is looking for files created by the previous profile (they catch up after an interation though). Between the two, I think your auto profile should work normally after this.

Also, if after doing the steps listed above, your miners appear not to be working in automatic profile, here is some troubleshooting help:
1) Detach from smartcoin, then run:

Code:

screen -r miner

You should connect to a screen session that has a tab for each instance which you can look at to see if the miners are actually running

2) Look in /tmp/ for smartcoin* files. These are the files created from the individual miner instances. You should make sure that they exist, and you can even look at the files themselves to make sure they contain the miner output.

Let me know what you find!

tigereye

member

Activity: 79

Merit: 10

Quote from: jondecker76 on July 16, 2011, 08:37:30 AM

tigereye

I just committed a change regarding the /tmp directory, perhaps it will help. smartcoin is meant to be installed and run as user. Here is what I would do:

1) Under user, make sure you clear the data directory:

Code:

rm -r ~/.smartcoin

2) Clear out any temporary data:

Code:

sudo rm /tmp/smartcoin*

3) Follow the install stuff again. Remove the old smartcoin directory and start again.

4) During the detection of the AMD/ATI SDK detection, there is a bug on some systems which shows two paths on one line. Since Linuxcoin is 64 bit, make sure to leave only the 64 bit relevant path!

5) If after starting things appear not to run, detach from smartcoin (option 3 on the control screen), then run

Code:

screen -r miner

This will bring you to the individual miner instances where you can see directly the error output of phoenix

Jon,

I've reinstalled smartcoin from scratch and re-setup all of my workers.
I was seeing the same symptoms as before (a bunch of "grep" error messages flying across the screen every few seconds, and no miners would stay up)

Then I created a manual profile and selected it, and it works!
I went back to the automatic profile, and it showed the same symptoms.

It seems something's funky with the 'automatic' profile.

Hope this helps you zero in on things. If you need any debug info to help narrow it down, let me know.

jondecker76

full member

Activity: 238

Merit: 100

Thanks for the reports guys! I believe that I have fixed the stable update bug (yes, the stable version was actually 425- sorry!)

Update r452(stable) and r453(experimental) available!

r452(stable/experimental):
- Lockup detection is now a lot less sensitive. Lockup detection works by comparing the output (the text on the entire screen) of the actual miner between each iteration. If its exactly the same, then a counter is incremented. The counter is reset each time the output looks different between iterations. It used to have to see 10 consecutive iterations to figure that the card associated with that instance was locked up. This turned out to be a bit too sensitive, as some cards run so stable that even the MHash reading doesn't change often. Also, while changing profiles (automatically or manually), there is a small amount of time where the miner screens don't change at all while it starts up. This value is now set to 50 iterations, which is probably a bit of an overshot (over 5 consecutive minutes where the miner screen doesn't change one bit), but at least it won't take people's miners offline on a false lockup.

- The stable update bug is fixed. It appears that the bug was introduced when the breakpoints were added. We will see with newer updates if the stable and experimental versions split again like they are supposed to over the next few updates.

r453(experimental):
- Failover detection made a little less sensitive. 10 iterations of no hashing is now needed to flag a profile as down (it was previously 8 iterations).
- This update was also made so that we can test that its only available to the experimental branch, and the stable branch should stay at r452

Future updates will have custom settings for the failover detection and deadlock detection.

Fletch

full member

Activity: 168

Merit: 100

I'll have a steak sandwich and a... steak sandwich

Quote from: plantucha on July 17, 2011, 03:57:21 PM

also I personally do not understand, why people pushing sqlite.
This will be big project and i can see people using it in mining farms with 50+ computers.
Mysql is perfect solution and for mailing postfix.
Think BIG.

Because SQLite is the perfect solution for something of this size (even when using 50+ computers). I've run a discussion forum with hundreds of concurrent users and several hundred thousand posts off SQLite. The issue with SQLite and Smartcoin is solved. No need to drag a bloated project like MySQL into this.

Quote from: plantucha on July 17, 2011, 03:57:21 PM

Safety first and what is better an more often maintained than Mysql or Postfix?
Some shady project without security upgrades?

LOL

plantucha

newbie

Activity: 56

Merit: 0

Quote from: xen82 on July 18, 2011, 06:44:16 AM

I started using smartcoin (r451 ex) recently and ran into an interesting bug: instead of failing over to a backup profile, smartcoin terminates with the following message:

07/18/11 14:05:06 ERROR: It appears that one or more of your devices have locked up. This is most likely the result of extreme overclocking!
07/18/11 14:05:06 It is recommended that you reduce your overclocking until you regain stability of the system
07/18/11 14:05:06 Killing Miners....

The thing is that it app fails at the very beginning when trying to connect to BTCguild using phoenix r111 while BTCguild is unaccessible. Perhaps this is because the meter doesn't show up until a connection is established? I'm running r451 experimental. Overclocking is not an issue here as everything is pretty much stock, with just my fans cranked up to keep everything as cool as possible.

Love your the work!

Yes me too.
around 5am EST BTCguild has problem with connectivity
=> failover switched to profile 1 and 2

all GPUs was already working on profile2 ( I have seen accepted work, so No GPU locked)
while profile 1 (BTCguild) was trying to connect and some GPUs had work, some GPUs was down.

then Anti-locking kicked in and killed all miners
before total switch to profile 2
and again 5 times after i start manually smartcoin

I have to set profile 2 without Failover to keep mining.

Overclocking is not a problem because my issues before I'm already on stock clock. using phoenix r108 + smartcoin r451ex

We need lower sensitivity overclock anti-lock.
or even better make anti-lock optional.

xen82

newbie

Activity: 22

Merit: 0

I started using smartcoin (r451 ex) recently and ran into an interesting bug: instead of failing over to a backup profile, smartcoin terminates with the following message:

07/18/11 14:05:06 ERROR: It appears that one or more of your devices have locked up. This is most likely the result of extreme overclocking!
07/18/11 14:05:06 It is recommended that you reduce your overclocking until you regain stability of the system
07/18/11 14:05:06 Killing Miners....

The thing is that it app fails at the very beginning when trying to connect to BTCguild using phoenix r111 while BTCguild is unaccessible. Perhaps this is because the meter doesn't show up until a connection is established? I'm running r451 experimental. Overclocking is not an issue here as everything is pretty much stock, with just my fans cranked up to keep everything as cool as possible.

Love your the work!

dishwara

legendary

Activity: 1855

Merit: 1016

Which is REALLY stable version?

plantucha

newbie

Activity: 56

Merit: 0

Quote from: jondecker76 on July 17, 2011, 05:43:19 PM

Update r450 experimental available
- Failovers can happen now in the case where you get unusually high rejected shares (over 10%)
- Dead GPU detection - detects locked up GPU (most likely cause is too aggressive of an overclock).
- AutoDonate bug fixed. I was finally able to reproduce it locally after a user PM'd me some good info. It only effected a few, and it took a specific combination of start time and donation minutes.
- The autodonate countdown timer is now displayed while donating again, and its accurate. I ran it through a battery of tests manually to ensure accuracy (tests included start/stop times within the same hour, start/stop times that crossed over an hour, and start/stop times that crossed over midnight)

elrock - do another update, you must have ran the update right inbetween my commits!

Update r450
- Dead GPU detection
I got false detection. 3+ hours stable running and smartcoin shutdown.
10min after I find out about it and just start smartcoin again.
Previous.log says shutdown because too much overclocking and GPU locked.
But GPUs are not locked all GPUs happily mining again without PC restart.

I'm overclocking GPU just about 6% over, my mem is factory set .
Temperatures 70°C, 70°C, 71°C, 72°C
All fans 100%
And it is tested stable for many days.

Now I'm on r451.

edit: when I wanna set back to factory speed
GPU 0 was showing error.
So it was locked.
Stats showing GPU in work confused me. Huh

Usually after smartcoin restart it shows DOWN.
MY BAD. Sorry for useless report. Cry

jl

newbie

Activity: 10

Merit: 0

plantucha-

Quote

just update catalyst to 11.6 it is easy copy/paste job, 10 min max:

that did the trick - now I don't need to rely on AMDOverdriveCtrl anymore, thanks! Have applied `--od-commitclocks` so my setclocks persists. ty

Rob P.

member

Activity: 84

Merit: 10

Quote from: jondecker76 on July 17, 2011, 06:19:48 PM

Quote from: elrock on July 17, 2011, 05:48:30 PM

Update isn't working for me -- it says the current stable version is 425 (I'm now on r449). Do I need to completely uninstall and reinstall? And is there a way to reinstall without wiping out my database?

How did you get to r449 if you're running stable?

You can go into Edit Settings, then change the development branch from 'stable' to 'experimental', after which time experimental updates will be available to you

Jon, something is up with the update system, see this:

Quote

smartcoin Management System r442(stable) Sun Jul 17 20:34:40 EDT 2011
--------------------------------------------------------------------------------
Getting svn information. Please be patient...
Your current version is r442(stable).
The current experimental version is r451
The current stable version is r425
Are you sure that you wish to perform an update?

You can see I'm running r442(stable), yet it's saying the current stable version is r425. Said the same thing when I actually updated, but I ended up with r442(stable).

jondecker76

full member

Activity: 238

Merit: 100

Quote from: plantucha on July 17, 2011, 03:57:21 PM

Quote from: jondecker76 on July 17, 2011, 06:36:58 AM

plantucha -

I'm just testing the cgminer and poclbm support.
So far for me, cgminer has been the slowest of the bunch for me (about 1040 MHash/sec vs about 1150 MHash/sec in phoenix). Here is the launch string that I have been testing with:

Code:

./cgminer -a 4way -t 2 -g 2 -o http://pit.x8s.de:8337 -u <#user#> -p <#pass#> -I 14

Also, I don't get much more than 0.7MHash out of my GPU no matter what I do (though, it is an older P4)

Let me know if you have any suggestions! I'll update this post once I can confirm that the cgminer support in smartcoin is complete

Maybe because cgminer has implemented
- Dynamic intensity that keeps desktop interactive under load and maximises throughput when deskop idle
so if you fiddling with PC intensity going down

or it can be just Intel/AMD sensitive

but anyway, I'm using CGminer for CPU only, Phoenix for GPUs

Code:

smartcoin Management System r442(experimental) Sun Jul 17 16:19:33 EDT 2011
--------------------------------------------------------------------------------
Host: localhost
GPU[0]: Temp: 69.00 load: 99%
GPU[1]: Temp: 70.00 load: 99%
GPU[2]: Temp: 71.00 load: 99%
GPU[3]: Temp: 72.00 load: 99%
CPU Load : 0.88%

Profile: Failover
--------BTCGuild--------
GPU[0]: [207.52 Mhash/sec] [132 Accepted] [1 Rejected] [RPC (+LP)]
GPU[1]: [207.48 Mhash/sec] [113 Accepted] [0 Rejected] [RPC (+LP)]
GPU[2]: [207.58 Mhash/sec] [132 Accepted] [1 Rejected] [RPC (+LP)]
GPU[3]: [207.51 Mhash/sec] [132 Accepted] [1 Rejected] [RPC (+LP)]
CPU:    [20.3 MHash/sec] [8 Accepted] [0 Rejected]
Total : [850.39 MHash/sec] [517 Accepted] [3 Rejected] [.580% Rejected]

Grand Total: [850.39 MHash/sec] [517 Accepted] [3 Rejected] [.580% Rejected]

Please note: for some reason after CPU mining is each GPU 2Mhash/s faster than without CPU mining.

now just tiny things like CPU load is 0.88% for GPU mining.
most likely 100% for CPU mining.

I'm getting CPU temperature from "lm-sensors", some people don't have lm-sensors installed.

also I personally do not understand, why people pushing sqlite.
This will be big project and i can see people using it in mining farms with 50+ computers.
Mysql is perfect solution and for mailing postfix.
Think BIG.
People don't cry for 50MB on your 1TB HDD.
Safety first and what is better an more often maintained than Mysql or Postfix?
Some shady project without security upgrades?

jondecker76 Thank you for your super fast development speed.

_______________________________________________________________________________ __________

P.S.:
I'm thinking about writing some documentation, soon it will be too complex for people who starting from this point.
But maybe after some front page simplification.
For example:
Pool configuration on front page
- it is already pre-configured, but users will try to do something just because they see it...like jen
I think first page have to be just this:

1) Reboot Computer
2) Kill smartcoin (exit)
3) Disconnect from smartcoin (leave running)
4) Select Profile
5) Configure
6) Update Smartcoin

and most important info:
Ctrl + a = switch to rigs live streaming
Ctrl + a +d = disconnect from smartcoin (leave running)

Submenu 5.

Mandatory basic Configure:
1) Configure Workers

Advacend configure:
2) Edit Settings
3) Remote computers: (yes/no) , address conf. etc

Profile configure:
3) Configure Profiles
4) Set Profile Failover Order

Pre-configured for basic use (tune for your own hardware):
5) Configure Miners
6) Configure Devices
7) Configure Pools

Now, I can see I have to more think about it. Because I'm not sure if is it much easier or more clear than before.
hmmm
....But at least now you do have idea what I mean.

Thanks for all the feedback! I do think that in the future, a new menu layout will need done. This project has gotten quite big and complex. I'll definitely take your suggestions into consideration!

Also, it would be great if you wanted to lead the charge for better documentation (sadly, I hardly have time to develop like I would like to, so updates in documentation have fallen behind). Perhaps starting a wiki entry would be a good starting point?

elrock

newbie

Activity: 41

Merit: 0

Quote

How did you get to r449 if you're running stable?

You can go into Edit Settings, then change the development branch from 'stable' to 'experimental', after which time experimental updates will be available to you

I don't know, but that's the stable version number that was available for update. I just updated to the experimental version to get rid of the error messages. Is it possible to revert to the last stable version, or am I stuck with experimental now?

jondecker76

full member

Activity: 238

Merit: 100

Quote from: elrock on July 17, 2011, 05:48:30 PM

Update isn't working for me -- it says the current stable version is 425 (I'm now on r449). Do I need to completely uninstall and reinstall? And is there a way to reinstall without wiping out my database?

How did you get to r449 if you're running stable?

You can go into Edit Settings, then change the development branch from 'stable' to 'experimental', after which time experimental updates will be available to you

plantucha

newbie

Activity: 56

Merit: 0

Quote from: jl on July 17, 2011, 05:28:12 PM

if I don't use AMDOverdriveCtrl - I get the following message after passing `aticonfig --od-setclocks=985,285 --adapter=all`:

Code:

ERROR - Set clocks failed for Adapter 1 - ATI Radeon HD 5800 Series  
        Please check that input values were valid

So AMDOverdriveCtrl must be doing something...And it's not `aticonfig --od-enable --adapter=all` ... that didn't change anything.

It would be good to know how to script Screen anyway - for general Linux education.

yes, ubuntu an other distros are using catalyst 11.5 -
yes you need AMDOverdriveCtrl if is your driver not up to date.
yes I did same , I was assuming I have 11.6.

just update catalyst to 11.6 it is easy copy/paste job, 10 min max:

http://wiki.cchtml.com/index.php/Main_Page

for ubuntu:
http://wiki.cchtml.com/index.php/Ubuntu_Natty_Installation_Guide

elrock

newbie

Activity: 41

Merit: 0

Update isn't working for me -- it says the current stable version is 425 (I'm now on r449). Do I need to completely uninstall and reinstall? And is there a way to reinstall without wiping out my database?

jondecker76

full member

Activity: 238

Merit: 100

Update r450 experimental available
- Failovers can happen now in the case where you get unusually high rejected shares (over 10%)
- Dead GPU detection - detects locked up GPU (most likely cause is too aggressive of an overclock).
- AutoDonate bug fixed. I was finally able to reproduce it locally after a user PM'd me some good info. It only effected a few, and it took a specific combination of start time and donation minutes.
- The autodonate countdown timer is now displayed while donating again, and its accurate. I ran it through a battery of tests manually to ensure accuracy (tests included start/stop times within the same hour, start/stop times that crossed over an hour, and start/stop times that crossed over midnight)

elrock - do another update, you must have ran the update right inbetween my commits!

elrock

newbie

Activity: 41

Merit: 0

I just updated to r449 and I'm getting a bunch of syntax error messages:

Code:

/home/elrock/smartcoin/smartcoin_ops.sh: line 494: local: `68': not a valid identifier
seq: missing operand
Try `seq --help' for more information.
expr: syntax error
expr: syntax error
/home/elrock/smartcoin/smartcoin_ops.sh: line 494: local: `1554': not a valid identifier
seq: missing operand
Try `seq --help' for more information.
expr: syntax error
expr: syntax error
/home/elrock/smartcoin/smartcoin_ops.sh: line 494: local: `25': not a valid identifier
seq: missing operand
Try `seq --help' for more information.
expr: syntax error
expr: syntax error

I froze it in mid-stream so there are probably more error messages along these lines. Shares are still being accepted, though.

Rob P.

member

Activity: 84

Merit: 10

Quote from: jl on July 17, 2011, 05:28:12 PM

if I don't use AMDOverdriveCtrl - I get the following message after passing `aticonfig --od-setclocks=985,285 --adapter=all`:

Code:

ERROR - Set clocks failed for Adapter 1 - ATI Radeon HD 5800 Series  
        Please check that input values were valid[\code]

So AMDOverdriveCtrl must be doing something...And it's not `aticonfig --od-enable --adapter=all` ... that didn't change anything.

It would be good to know how to script Screen anyway - for general Linux education.

You don't want AMDOverdriveCtrl commands running in Screen. Screen is for running commands that KEEP RUNNING and you then want to detach from them, but keep them running. The AMDOverdriveCtrl commands just run and exit. Screen isn't necessary.

However, here is an example of running something in screen:

Code:

/usr/bin/screen -dmS gpu0 /usr/local/miners/startminer.sh 0

That will execute the script "startminer.sh" and pass it a "0" parameter. It will name the screen "gpu0". It will immediately detach from the session, so on the command line it won't look like anything happened.

To attach to the screen you'd type:

Code:

screen -r gpu0

To detach from the screen again, you'd type CNTRL-A, D

jl

newbie

Activity: 10

Merit: 0

if I don't use AMDOverdriveCtrl - I get the following message after passing `aticonfig --od-setclocks=985,285 --adapter=all`:

Code:

ERROR - Set clocks failed for Adapter 1 - ATI Radeon HD 5800 Series  
        Please check that input values were valid

So AMDOverdriveCtrl must be doing something...And it's not `aticonfig --od-enable --adapter=all` ... that didn't change anything.

It would be good to know how to script Screen anyway - for general Linux education.

Rob P.

member

Activity: 84

Merit: 10

Quote from: jl on July 17, 2011, 05:01:47 PM

Quote from: jondecker76 on July 15, 2011, 09:23:05 AM

If there are any commands that you want to have run automatically when smartcoin starts, create a file "init.sh" in the smartcoin directory. Then you can add your commands that you want run on startup.

Admittedly undereducated about exactly what AMDOverdriveCtrl is doing to my overclock settings/performance - but I have to start it for both cards everytime I reboot my rig. How can I script this with a detached screen? I'm trying:

Code:

#!/bin/bash
screen -dmS ctl0 "AMDOverdriveCtrl -i 0"
screen -dmS ctl1 "AMDOverdriveCtrl -i 3"

Put the commands you want in init.sh in the smartcoin directory and they'll run every time you start smartcoin.

Topic: Smartcoin Linux mining administration. [MULTI-MACHINE SUPPORT NOW IN!] - page 16. (Read 105029 times)