BAMT version 0.5 - Easy USB based mining Linux with farm wide management tools

[mc_lovin]

Code:

root@bamt-miner:~# rm /home/user/.Xauthority
rm: cannot remove `/home/user/.Xauthority': No such file or directory

doesn't seem to be there.. but if we can figure out how to reset Xauthority manually like this, I'll be psyched!

[1715036604]

1715036604

[1715036604]

1715036604

[gnar1ta$]

Code:

root@bamt-miner:~# rm /home/user/.Xauthority
rm: cannot remove `/home/user/.Xauthority': No such file or directory

doesn't seem to be there.. but if we can figure out how to reset Xauthority manually like this, I'll be psyched!

As root:

Code:

xauth merge /home/user/.Xauthority

[pehoko]

So problems continues after new unexpected power down : xauth:  error in locking authority file /root/.Xauthority
message : read-only file system
Starting mining processes...: minestart_mining: starting mining processes
..muninrm: cannot remove `/etc/munin/plugins/gpuhash_all': Read-only file system
rm: cannot remove `/etc/munin/plugins/gputemp_all': Read-only file system
rm: cannot remove `/var/lib/munin/plugin-state': Read-only file system
ln: creating symbolic link `/var/lib/munin/plugin-state/run': File exists
..GPU 0..fan 0..OC 0..GPU 1..fan 1..OC 1..GPU 2..fan 2..OC 2.

I cant reboot machine, or start mining  

usb key has died.  might work if you reflash, might not, or might work for a day or two and die again.

Yes you were right, it was active for 20 hours after clean format and again this error. Strange but how it dies:
http://img600.imageshack.us/img600/4928/adata.jpg

[BitMinerN8]

So problems continues after new unexpected power down : xauth:  error in locking authority file /root/.Xauthority
message : read-only file system
Starting mining processes...: minestart_mining: starting mining processes
..muninrm: cannot remove `/etc/munin/plugins/gpuhash_all': Read-only file system
rm: cannot remove `/etc/munin/plugins/gputemp_all': Read-only file system
rm: cannot remove `/var/lib/munin/plugin-state': Read-only file system
ln: creating symbolic link `/var/lib/munin/plugin-state/run': File exists
..GPU 0..fan 0..OC 0..GPU 1..fan 1..OC 1..GPU 2..fan 2..OC 2.

I cant reboot machine, or start mining  

usb key has died.  might work if you reflash, might not, or might work for a day or two and die again.

Yes you were right, it was active for 20 hours after clean format and again this error. Strange but how it dies:
http://anonymouse.org/cgi-bin/anon-www.cgi/http://img600.imageshack.us/img600/4928/adata.jpg

Have you tried a "new" USB flash drive?

[pehoko]

No, I do not have "new" ... using old hard disk again.

[-ck]

A bit later, gigavps found that 2.3f would sometimes just exit with no clear indication of why.  He was desperate for a fix, and I believe contacted you guys, posted on the cgminer forum, etc.  When no solution was forthcoming, he asked me if I could hack mother to detect that it had exited and start it back up.  So I did.

Anyway I instrumented the newer cgminer restart problem in bamt and it is indeed in the /opt/bamt/mother monitoring file. The perl parsing of ps axu output fails to detect cgminer is there and it then calls /etc/init.d/mine restart . So it is in fact bamt that is restarting cgminer, and if I edit the mother script to parse the output of ps axu differently, it does NOT continually restart cgminer. Alas I don't really know perl, but it does seem to need fixing.

If you copy this file on top of mother in /opt/bamt it shouldn't restart on you (note this is not the correct fix and will probably not restart cgminer should it actually really crash):
http://ck.kolivas.org/apps/cgminer/temp/mother

So what would be best thing to do then, use ckolivas's fix OR don't apply BAMT fix 20 in the first place?

fix 20 is what causes this problem in the first place. It's mean to be a watchdog for when cgminer fails but on cgminer 2.6+ it actually kills cgminer instead. If you want to use the latest cgminer, you either don't apply fix 20, or use my temporary fix above if you've already applied it.

[philips]

Got it, thanks.

[lodcrappo]

Two identical boxes, both MSI 890FXA-GD70 motherboards, both 4x 5850s, I cannot set the password without it spitting the Xauthority error.

We need 6 cards running on each box.  It's like effin' impossible to do but there is a 6-card 890FXA-GD70 with 6x 6870s on it running no problem.

These two, again identical, all same bios (1.6) and it errors on 4 cards.  WTF.

Here is what happens at boot..

1 - normal linux stuff
2 - x windows config blown away
3 - ati device detection -> generate new x win config
4a - x starts in one thread
4b - mine_start starts in another thread, sleeps for start_delay
5 - X finally gets going
6 - default user logs in automatically to x session
7 - mine_start script runs xauth + to make the X server allow connections

now, in root's .bashrc there is a command: xauth merge /home/user/.Xauthority
this runs when you log in as root.  it gives root the ability to talk to the X server.

however, if you log in via ssh before step 6 completes, the .Xauthority file isn't there yet and you get no joy.
or if step 7 happens before step 5, you'll get no joy.

some USB keys are much, much slower than others.  same with cpu, etc.  having more GPUs means step 3 can take much longer.  some GPUs seem to make step 4a/5 take a really long time if no monitor is attached.  there are many variables here.  since the process splits in step 4, there isn't a strong tie between them and things can get out of order, or you can just be in a rush and logging in via ssh too soon.

the default timing works in most situations, but if it doesnt your best bet is to connect a monitor and change the root password using the console, after you can see that everything is started up.  open a root terminal and do passwd.  once you've done that, it will probably work fine.  if not, you may need to increase the startup delay, its the setting "start_mining_init_delay:" in the main section, see http://bamter.org/redmine/projects/bamt/wiki/Examples

[mc_lovin]

Code:

root@bamt-miner:~# rm /home/user/.Xauthority
rm: cannot remove `/home/user/.Xauthority': No such file or directory

doesn't seem to be there.. but if we can figure out how to reset Xauthority manually like this, I'll be psyched!

As root:

Code:

xauth merge /home/user/.Xauthority

Thanks for this!  Definitely handy to know the command now, I'll save it in my notes!  It does just give me the same error again, but at least I can call the error now.

Two identical boxes, both MSI 890FXA-GD70 motherboards, both 4x 5850s, I cannot set the password without it spitting the Xauthority error.

We need 6 cards running on each box.  It's like effin' impossible to do but there is a 6-card 890FXA-GD70 with 6x 6870s on it running no problem.

These two, again identical, all same bios (1.6) and it errors on 4 cards.  WTF.

Here is what happens at boot..

1 - normal linux stuff
2 - x windows config blown away
3 - ati device detection -> generate new x win config
4a - x starts in one thread
4b - mine_start starts in another thread, sleeps for start_delay
5 - X finally gets going
6 - default user logs in automatically to x session
7 - mine_start script runs xauth + to make the X server allow connections

now, in root's .bashrc there is a command: xauth merge /home/user/.Xauthority
this runs when you log in as root.  it gives root the ability to talk to the X server.

however, if you log in via ssh before step 6 completes, the .Xauthority file isn't there yet and you get no joy.
or if step 7 happens before step 5, you'll get no joy.

some USB keys are much, much slower than others.  same with cpu, etc.  having more GPUs means step 3 can take much longer.  some GPUs seem to make step 4a/5 take a really long time if no monitor is attached.  there are many variables here.  since the process splits in step 4, there isn't a strong tie between them and things can get out of order, or you can just be in a rush and logging in via ssh too soon.

the default timing works in most situations, but if it doesnt your best bet is to connect a monitor and change the root password using the console, after you can see that everything is started up.  open a root terminal and do passwd.  once you've done that, it will probably work fine.  if not, you may need to increase the startup delay, its the setting "start_mining_init_delay:" in the main section, see http://bamter.org/redmine/projects/bamt/wiki/Examples

Yeah I put the start_mining_init_delay on all the boxes, it just seems like a good idea to put that there.

Interestingly, I can't connect a monitor to change the root password using the console.  When I have 5 cards on there, I can, but if I put the 6th on, there is absolutely *no* video, not even a post...  But the LED blinks, and the system boots, and I can log in via SSH but it can't merge the Xauthority.  Clearly if the thing can't light the screen up at all, that tells us something is wrong, and I'm positive it's hardware...  We're asking too much of these GD70's.  Having two identical boards, identical BIOS's, and identical settings and one will run 6 cards and the other will not?  Last night I came up with a theory: the box that can run 6 cards is running 6x 6870s, where 6x 5830s or 5850s is a no go, but 5 is.  Now that this has been brought to my brain's attention, they are not as identical as I was thinking.  Now I want to experiment with other BIOS versions, I figured if v1.6 can run 6 cards, it is the one to go with, but maybe another version can run 6x 5830s?  To be continued!

Thanks for your troubleshooting! 

[philips]

Have you thought about copying the .Xauthority file from one rig that has it to all the others that don't?

[mc_lovin]

Have you thought about copying the .Xauthority file from one rig that has it to all the others that don't?

Hmm..  Yea I think the problem has been somewhat solved.  After way too long of configuration switching and trying different everythings, totally pulling my hair out, I decided to set up a workbench and test each and every card on their own.  There was a card in each of the two machines that were bunk, they would show up in linux and even mine but they did not display video on POST or anything if they were the primary card. 

Yeah, that would probably cause an Xauthority issue, no?  /facepalm

[pehoko]

In my practice each card is tested individually, then flash bios and after this is mounted on the rig.  I saw this error after new rig starting and maybe the problem was sticky PCE-eX1 extender. X server didn't start - waiting count and idle.

[edimode]

I see people keep asking for a BAMT able to run with 7970 cards so I am thinking to share mine.
It may work for you or it may not...give it a try. It worked for me two months already so is not gonna blow your PC.

It is a BAMT updated to SDK 2.6 and 8.921 AMD driver using cgminer 2.5.0.
SDK 2.6 was recommended by ckolivas for 7970 and the 8.921 driver was (I tried all recent drivers) the only one that did not gave me hardware errors with cgminer.
Other (more recent) SDK/driver combinations could work also but this one worked perfectly for me.

All you need to do is edit bamt.conf at the cgminer_opts line and enable cgminer in each GPU section for how many cards you have.
If you want to use a separate cgminer config be my guest, I personally enter all the options I need in that cgminer_opts line.

I give no guarantees that it will work for you, it does for me with three 7970 cards.

root password = root
user password = live

Download:
https://ca.isohunt.com/torrent_details/402809617/BAMT+for+7970?tab=summary

Hope this damn torrent will work, eventually add these trackers to your bit-torrent client:

http://announce.opensharing.org:2710/announce
http://exodus.desync.com:6969/announce

It works. Thanks a lot.
But, by default in this modded image GPU0 enabled with frequency = 1150Mhz and this is to much for hd7950.

My rig setup:
5x Sapphire HD7950 (SKU ...01-40G)
PSU: Aerocool Strike-X 1100W
MB: Asrock 970 Extreme 4
2GB RAM
4x 1x-16x risers

Sorry for my english.

P.S. I'm seeding this torrent too.

[philips]

It works. Thanks a lot.
But, by default in this modded image GPU0 enabled with frequency = 1150Mhz and this is to much for hd7950.

Glad to hear.

1150 was just a safe value one can use with 7970, (actually I use higher frequencies: 1185/1190) and the whole cgminer line in bamt.conf is just a template, you must modify it (intensity, memclock, powertune etc) according to your cards.
To be honest I only thought about 7970 cards, not 79xx series in general.

Cheers

[yochdog]

So BAMT still does not support 7970's straight outa the box? 

[philips]

No, the official one does not.
But I think lodcrappo is cooking something big for the next release.

[abracadabra]

So BAMT still does not support 7970's straight outa the box? 

a little birdie told me .6 should include 7xxx support

[yochdog]

daaaaamit!

[yochdog]

So BAMT still does not support 7970's straight outa the box? 

a little birdie told me .6 should include 7xxx support

any word on release date?

[TheHarbinger]

I am constantly amazed by the inability of people to do something so simple as update the SDK on a Linux machine. 

A few minutes on a search engine, and a few minutes using the shell, and you could have been using 7xxx cards months ago.