[Bounty 0.1 BTC] MinePeon stability issues

[LogicalUnit]

Hi all,

I've been using MinePeon 0.2.3a for a few days, and have fixed one stability issue related to the system clock not synchronising over wifi. See this post: https://bitcointalk.org/index.php?topic=224015.msg3086667#msg3086667

However, both MinePeon and cgminer seem to be restarting frequently, giving me only a few hours of uptime between cycles.

I know that the first step would be to check the MinePeon logs, but I don't know where they are saved.

Any help would be much appreciated.

Thanks

[1715277181]

1715277181

[LogicalUnit]

Added bounty

[LogicalUnit]

Right, I have followed these instructions to upgrade to cgminer 3.4.2 and we shall see if that improves anything.

[Scorpia]

also you might want to look at updating the http git.

as for the time, check http://www.minepeon.com/forums/viewtopic.php?f=15&t=95

it allows the time to be stared on power outage so you dont need to wait for the ntp to sync.

as for general 0.2.3 stability. i found i got much better stability when i went back to 0.2.2 then did the git pull.

you end up with the same minepeon but it seems like a more reliable arch linux.

[LogicalUnit]

also you might want to look at updating the http git.

Not sure how to do this. . .

[Scorpia]

try this

cd /opt/minepeon/http
git pull

then I would suggest a

sudo reboot

[LogicalUnit]

Okay, I have reverted to MinePeon 0.2.2 with the latest ntp, cgminer and http packages. I'm trying to run the status screen as described here but the script terminates and the screen stops updating when I disconnect my ssh client. How do I make a python script continue after disconnecting ssh?

[LogicalUnit]

Code:

sudo python2 cgminerLCDStats.py &

The ampersand did the trick

Now to see how stable it is. . .

[LogicalUnit]

14 hours of uptime so far. Excellent!

[tk1337]

I know that the first step would be to check the MinePeon logs, but I don't know where they are saved.

Not sure if you saw my response in the MinePeon thread, but the system log files are located:

Code:

/var/log/

sudo cat /var/log/auth.log
sudo cat /var/log/crond.log
sudo cat /var/log/daemon.log
sudo cat /var/log/errors.log
sudo cat /var/log/everything.log
sudo cat /var/log/faillog
sudo cat /var/log/kernel.log
sudo cat /var/log/lastlog
sudo cat /var/log/messages.log

(Use sudo cat to output the entire contents of the file or you could do sudo tail -n [number] /var/log/[filename].log (Example: sudo tail -n 50 /var/log/auth.log would give me the most recent 50 entries within that file.)

auth.log = authorization systems log
crond.log = crontab log / cron jobs
daemon.log = system & application daemons (usually background processes)
errors.log = system errors.
everything.log = *should be **everything*** (hehe)
faillog = login failure log.
kernel.log = messages from the linux kernel
lastlog = login log
messages.log = general messages and system notices.

[LogicalUnit]

Very helpful. Many thanks.

[LogicalUnit]

Hmm, now my LCDStats script has stopped working!! I'm getting the following output:

Code:

Welcome to cgminerLCDStats
Copyright 2013 Cardinal Communications
Main Exception Handler:
Failed to claim interface

Traceback (most recent call last):
  File "cgminerLCDStats.py", line 473, in <module>
    displayErrorScreen(str(e))   # something bad happened, better display the error screen
  File "cgminerLCDStats.py", line 198, in displayErrorScreen
    display = LCDSysInfo()
  File "/home/minepeon/cgminerLCDStats/pylcdsysinfo.py", line 332, in __init__
    raise IOError("Failed to claim interface")
IOError: Failed to claim interface

[Zeek_W]

Have you tried imaging the SD card to a new one? Not sure at all if it is the problem - but they do have a finite life.

[LogicalUnit]

The SD card is only a few weeks old. I have another one I could use, but I'm confused to why the LCD screen suddenly stopped working.

[LogicalUnit]

Hmm, after a few power cycles, LCDStats is working again. No idea what was causing it to fail. After investigation, MinePeon 0.2.2 and cgminer 3.4.2 were running properly, it was just my LCD that froze, leading me to believe something else was wrong.

[tk1337]

I got this issue yesterday, tbh, it just means it couldn't locate the lcd display from the GPIO.

I'm starting to think a longer delay period before running CGMiner on bootup is a really great idea. As someone mentioned in the other post the factor of time (since there's no hardware clock) can skew CGMiners stats and more-than-likely cause a few other issues, I would imagine. So, I'm going to write a bash script to stage things, start processes in an order, then (if it's actually starting a process that should be constantly running) check said process to make sure it's running before starting the next item. I haven't really taken the time to look to the current script that starts CGMiner other than a few quick glances.

Granted on a bootup CGMiner might take a few to load, but I'd rather that then have random oddities.

[cardcomm]

I've duplicated this issue, and posted a "fix" for it here:

https://bitcointalk.org/index.php?topic=137934.msg3149341#msg3149341

In short, it's likely a bad USB cable.

[cardcomm]

Code:

sudo python2 cgminerLCDStats.py &

The ampersand did the trick

Now to see how stable it is. . .

BTW, I'm not actually running v0.2.3a. But I think the best way to start and stop cgminerLCDStats on that version is:

systemctl start cgminerLCDStats
systemctl stop cgminerLCDStats

I actually just updated the MinePeon readme on the cgminerLCDStats github about this.

[LogicalUnit]

Great work, mate. The little MtGox ticker is very cool

[LogicalUnit]

I've duplicated this issue, and posted a "fix" for it here:

https://bitcointalk.org/index.php?topic=137934.msg3149341#msg3149341

In short, it's likely a bad USB cable.

Hmm, I've tried 3 different USB cables, including crimping one, and none of them work. The device shows up when I do an lsusb as Van Ooijen Technische Informatica shared ID for use with libusb, but I get the same error.

[LogicalUnit]

. . .and now it's working again. Only thing I did was a sudo python2 cgminerLCDStats.py, followed by ctrl-c, and then sudo python2 cgminerLCDStats.py &

[cardcomm]

I've duplicated this issue, and posted a "fix" for it here:

https://bitcointalk.org/index.php?topic=137934.msg3149341#msg3149341

In short, it's likely a bad USB cable.

Hmm, I've tried 3 different USB cables, including crimping one, and none of them work. The device shows up when I do an lsusb as Van Ooijen Technische Informatica shared ID for use with libusb, but I get the same error.

PM'ed you with another potential fix

[viriat0]

Always had instability with version 0.2.3, changed to 0.2.2.

Uptime
0.2.2 cgminer3.3.1 3 days 7 hours 46 minutes 51 seconds 3 days 7 hours 46 minutes 13 seconds.

Setup
7 gizmos
D-linh hub7
Mitsai

Cheers

[cardcomm]

Did you get the LCD display working?

[LogicalUnit]

Did you get the LCD display working?

Yes! https://bitcointalk.org/index.php?topic=290511.msg3156977#msg3156977

However, the script seems to terminate after several hours, and I need to restart it. Could there be a memory leak or something in the code?

[LogicalUnit]

I'd also like to bring something to the attention of the MinePeon WebGui developers: when diff is greater than 1 in cgminer, the WebGui reports the number of accepted shares as the total accepted shares divided by the diff. So in my rig with 10 AMUs, cgminer 3.4.2 sets my diff to 10 automatically. This means that while one of my AMUs has 3400 accepted shares, only 340 accepted shares appear in the WebGui, skewing all of the statistics.

[tk1337]

I'd also like to bring something to the attention of the MinePeon WebGui developers: when diff is greater than 1 in cgminer, the WebGui reports the number of accepted shares as the total accepted shares divided by the diff. So in my rig with 10 AMUs, cgminer 3.4.2 sets my diff to 10 automatically. This means that while one of my AMUs has 3400 accepted shares, only 340 accepted shares appear in the WebGui, skewing all of the statistics.

Granted I stopped using CGMiner due to instability with working with over 20 miners, however, there are some stats still off from console to webUI (mainly HW Error % from my screenshots). I really don't think the code is doing anything with the stats from the API other than pulling them directly from the miner, but I am going to take a look into it, I did see a few things awhile back.

BFGMiner Console


BFGMiner WebUI

[kano]

You can't divide those two numbers and get a correct answer for anyone except a few.

[cardcomm]

Did you get the LCD display working?

Yes! https://bitcointalk.org/index.php?topic=290511.msg3156977#msg3156977

However, the script seems to terminate after several hours, and I need to restart it. Could there be a memory leak or something in the code?

Well, I can't preclude the chance of a memory leak, but my script runs days at a time without failing. Are you seeing an error message? I seem to recall you getting an error to the effect that it's unable to claim the USB device.

[LogicalUnit]

It seems to be claiming the USB normally, and it's been running for about a day now. I'll let you know how it goes.

[LogicalUnit]

Okay, the stats screen froze again in just under 2 days of mining (about 45 hrs of uptime). When I restarted the stats script, I noticed that cgminer had been restarted as well, a few hours earlier. This implies that something in the system is causing problems for both.

Can someone help me debug?

[LogicalUnit]

Anyone? I'm still getting cgminer restarts and LCD script termination.

[tk1337]

Anyone? I'm still getting cgminer restarts and LCD script termination.

Switch to BFGMiner and see if that solves your issues, has for me and several others lately.

[LogicalUnit]

Anyone? I'm still getting cgminer restarts and LCD script termination.

Switch to BFGMiner and see if that solves your issues, has for me and several others lately.

Can you provide instructions?

[LogicalUnit]

Okay, I've got bfgminer up and running, but now my http stats don't work. Hrm. . .

[tk1337]

Okay, I've got bfgminer up and running, but now my http stats don't work. Hrm. . .

I saw your post in the other thread, tbh, I haven't ran the http from 0.2.2 in a long time now, so with that being said...
I can attempt to pull it on one of my rPi's and see what could be the issue for you (or) attempt to download the latest version of http (see my information about backing up your current http directory) however, I do know the latest of http 0.2.3a is still under dev. so you will probably run into issues there as well... I can't remember the simple fixes that were needed for the latest dev build, as I have so many random projects going on within my http directory for stuff I'm coding, I haven't done a pull in awhile.

But I can easily do an update, pull the http from 0.2.2 and see what the issue is, then tell you how to fix it (if I am indeed having the same issue). I'll do it on my other rPi that's just acting as a proxy for the blades, I never use the webStats on that one.

Sorry for not responding earlier about switching to bfgminer, haven't looked at the forums in a little bit...

Code: (Switch from CGMiner to BFGMiner on MinePeon)

sudo systemctl stop cgminer.service
sudo systemctl disable cgminer.service
sudo systemctl enable bfgminer.service
sudo systemctl start bfgminer.service

[LogicalUnit]

But I can easily do an update, pull the http from 0.2.2 and see what the issue is, then tell you how to fix it (if I am indeed having the same issue). I'll do it on my other rPi that's just acting as a proxy for the blades, I never use the webStats on that one.

Cool, thanks for checking it out