Topic: cgmon - mining monitor for Linux - auto restart, reboot, sick gpu, ASIC, &more - page 12. (Read 48345 times)

Yeah I'm doing a fresh install of everything and will see how it goes. I might even start from scratch with my settings. Hard hang is no fun. It can go 5 hours w/o hard hanging and then 5 times in 1 hour. Really hard to put your finger on.

From my experience, it can be due to:
1: Unstable thread-concurrency for the card
2: Too much OC
3: Too much undervolting
4: A combination of OC and it now needs more voltage
5: Somekind of other non-default value which is incompatible with your card, many errors are caused by manual settings!
6: Changes to engine/ memory clock values, or voltage value while running cgminer/sgminer can sometimes crash it, no idea why...

All of them will usually result in what is earlier mentioned as "ASIC hang", which in English means that the graphic driver has crashed.
That will either lead to "dead GPU", or a complete cgminer hang.

It could also be due to a unstable/uncompatible driver, but not sure that has caused me any issues.

I myself am running with a setup of:
Xubuntu 12.10
AMD Catalyst 13.12
AMD ADL SDK 6.0
AMD APP SDK 2.8
Sapphire 7950 w/ boost, running 1100/1500 at vddc 1.100, thead-concurrency 24000

Angela/bobers:
That is because of this line in the cron:
*/1 * * * *     /usr/local/bin/cgmon.tcl >>/var/log/cgmon.log 2>&1
By adding >>/var/log/cgmon.log I guess you override the logpath setting in cgmon. You should be fine by doing:
*/1 * * * *     /usr/local/bin/cgmon.tcl

A little tip: */1 is the same just *, so * * * * * is every minute.

Keep in mind that as there is no username set, it will run as currently logged on user. That tells me for the script to work, as it needs root rights, that you either:
1: Is always running as root, and thereby the miner process as root, generally not a good idea for security purposes
2: Not using the crontab file which is compatible with usernames
3: Not using the crontab file which is compatbile with usernames, and you are running with a non-root user account, and in that case the script won't work (reboot command requires it).

It's only in /etc/crontab you can specify run-as usernames, not by doing crontab -e, which is a seperate file for each user, and thereby always runs as that user.

So what you really should do is:
nano /etc/crontab
Add:
* * * * *   root   /home/username/cgmon.tcl

What are the possible causes of sgminer hard hang? I was getting GPU not well before and tuned it down and now sgminer just hangs.

Could you plz give me steps how did you add it to BAMT?

If set crontab like this
and edit cgmon.tcl like this
the log file like this

no big deal, but maybe fix the Install instructions ?
thanks for your gread job.

That file looked fine.  BAMT 1.3

Hmm.  Take a look at /tmp/cgmon-mine.sh.  That's the file that has the mining command.  It should also contain:

You could try running /tmp/cgmon-mine.sh directly and see if that gets you the full hashrate.  If it does, I expect the problem could be with the way cgminer is started from crontab.   What OS are you using?

Fixed.  Thanks!

Try changing:

"exec shutdown -r now"

To:
"exec reboot"

That should work in CentOS.

OS: CentOS 6.2 X64
cgmon 0.1b13



[root@rig ~]# whereis shutdown
shutdown: /sbin/shutdown /usr/share/man/man8/shutdown.8.gz /usr/share/man/man3p/shutdown.3p.gz /usr/share/man/man2/shutdown.2.gz

Thanks for the PHP time out catch.  That should do it.

When cgmon restarts cgminer, I find my hashrate is less than half what it was before, despite all the same settings.  Does the restart command not include (in windows terms) setx use_gpu_alloc 100 and the gpu_use_sync_objects 1?  Also, any word on the reboot if the hash rate falls below xx?

cgmon b12 released

http://www.forked.net/~apex/cgmon/cgmon.tcl

Changes:

# 0.1b12
#   Added check for connection reset when using cgminer API.
#   Moved share monitoring option into config section.
#   Changed pool configuration.  Now you can comment out unused pools or leave them blank.
#   Added option to disable email notifications when cgmon.tcl is run by hand.  Keeps the inbox spam down when adjusting settings, etc.
#   Cleaned up some code.
#   Changed cronjob default to every 2 minutes.  If you have an existing cronjob, you'll need to edit /etc/crontab yourself to make this change.

Actually it could be twice that much loss since cgmon.tcl only has a chance to catch the problem once every 5 minutes...  The next version will have a recommended cron time of every 2 minutes.

FYI, having your miner start four times a day, via a 5 minute cronjob would on average potentially waste 0.6% of your daily mining time, vs a 1 minute cronjob.    Perhaps I should change the default cronjob to 1, 2, or 3 minutes?

Of course the reboots themselves could cause that much loss too... 

The nice thing about using cgmon.tcl for me, is that I can see which GPU's are clocked too high and slowly lower those clock speeds until the GPU becomes stable over multiple days or weeks.

By default the cronjob runs every fifth minute of the hour.  So your miner would get started within 5 minutes of booting.  If you want, you can change */5 to */1 or */2 so it checks every minute or two instead.   

haha... yea.  That feature is *almost* working

Sweet implementation, much smoother than mine!

Now you can also cleanup 50 % of your todo!

Done.

#   Changed pool configuration.  Now you can comment out unused pools or leave them blank. 

Note: In 0.1b12 the pool variables have changed format so copy/pasting old pool info into this latest download will break the script.  You'll need to enter the actual address, username, and password as required.

Great point.  I will have to change it to allow that behavior.