Topic: Smartcoin Linux mining administration. [MULTI-MACHINE SUPPORT NOW IN!] - page 5. (Read 105029 times)

How do you update to experimental versions ?

Hey john

Have you been able to iron out the poclbm issue yet ?

Update r638 now available!
- Autodetection routine now assumes phatk kernel is available on phoenix installs (only ancient phoenix installs don't have the phatk kernel available).  The phatk kernel is now the default on any new auto-detected phoenix install (though you can allways go to Edit Miners from the control screen and change it to poclbm if you want)

- 'updatedb' should run correctly remotely now during autodetect (without tty errors)

- The hardstatus line at the bottom of the miner instances is back, and now even works on remote machines

- Some small low-level changes that let me add extra flags to the SSH communications for some commands.

Multi-Machine support release!

Update r636e now available!
- Not everything is done that I wanted, but multi-machine support is in!
- All communications are securely tunneled through SSH via uniquely generated keys (every installation has their own unique public/private keys generated)
- Added a 'Log' tab to the smartcoin screen session which tails the smartcoin.log to make it easier to see whats going on
- If you are running a custom lockup.sh script,  the machine number firing the lockup is passed as the first parameter so you can alert yourself of which machine caused the problem, take different actions based on machine number, etc...


KNOWN ISSUES WITH MULTI-MACHINE SUPPORT:
- during auto detection run, the updatedb command fails.  I'd recommend running it yourself on the remote machine before hand
- If a remote machine normally runs smartcoin as root (with sudo - Linuxcoin would be an example of this), then when setting up a remote machine from Configure Machines, you must use the username 'root'.  This also means that the remote machine must have a root password for this to work.  To set up a root password, on the remote machine (before setting up the machine entry on the control screen):
then enter a password twice.  You can now set up the remote machine in Smartcoin using the 'root' user

- When editing a machine, you can select to disable it.  The disabled flag DOES work, however a restart of smartcoin is required to see the disabled flag (I.e. if 'Machine#3' is running, and you go to Edit Machines and change its disabled flag to yes, then the status display for screen #3 will stay there until a restart of smartcoin is performed.

- Things aren't yet optimized on the multi-machine side of things, and a lot of live database queries are going on. This isn't an issue, but may show as a little extra CPU usage and or slow refreshes on the remote machines. Optimizations to this part of the code will happen once I see how it works for everyone.

- detected phoenix installations on remote machines default to poclbm kernel. You can Configure Miners->Edit for the machine, and change it manually to phatk if you want (will work on fixing soon)

- cgminer autodetection isn't working currently, but you can always add the entry manually for now.

- The multi-machine functionality was made with local LAN networks in mind, but it will work with a host across the internet - just don't expect blazing refresh speeds if you use a remote host across the internet.

 - The <<>> status message is disabled for now until I can make a multi-machine safe way to determine if a miner session launched successfully or not

A BIG BIG BIG thanks to EnzoMatrix, krzynek1 and plantucha for allowing me remote access to their machines.  I did some test on all of these machines and everything appears to be working.  I did my best to keep their downtime to a bare minimum



Please give it a try, and report your problems or successes. The sooner we verify how things work on different installations, the sooner I can get to optimizing and fixing the newest problems.

anybody tested
SDK 2.5 ?

AOCLBF https://bitcointalksearch.org/topic/aoclbf-was-phoenix-rising-phoenixpoclbm-gui-with-built-in-overclock-tool-10264 is also able to overclock/volt the Sapphire Xtreme 5850's. 
aticonfig works in linux.

Running my Xtreme's in the range of 950-990 on stock voltage.

As far as I have read, the Sapphire Xtreme 5850's do not allow you to Bios overvolt via RBE..  Trixx is the only reliable method.  I could be wrong.

If you are limited by software overclocks and want to do the same on linux, look at Radeon Bios edito (RBE) , Ive solved somewhat similar issues on linux by simply modding the bios of the cards directly.

I have not built it yet, (3 Cards are shipping today from NCIXUS, and 4 more from amazon later in the week) but I have seen reports of it being stable at 1.168v running over 400-440 mh/s.

YMMV so I want to test it.   

it has a good chance of working on cygwin,  you will need gnu screen working under cygwin though (I've read there is a patch).

what's your hashrate on your windows boxes?   I get a 100% stable 385 out of my 5850s under linux with no overvolt using 3 instances per card

Any idea if it would work (Remote Machine) using a SSH into a Windows Box running Cygwin? 

I am loathe to do it but it looks like I will need to install Windows on a couple of miners in order to overvolt my 5850's via Trixx.  If it wasn't a  difference of 500-700 mh/s over what I could get in linux I wouldn't even contemplate it.

Theoretically it is just starting the miner via a string command over ssh, so it should work....in theory.

Also will be posting an update to the PHP scripts I have now hopefully tonight.  Now I can monitor hashrates, accepts, rejects, per pool and profile via the smartcoin.xml in the /tmp folders via the web as well as changing profiles on the fly.

I haven't committed any of the new code yet, but just giving an update.

The multi-machine stuff is going very well!  I just finished implementing the Edit Machines part (complete with re-syncing connection keys if login information changes).  As of now, multi-machine support is working 100% as expected - tested on one machine.  I'm still looking for another machine or two to test it on if anyone can spare about half an hour or so of downtime.

As of right now, there is only one more feature to implement.  Currently, you can only "reload" all miners on all machines (after you make a settings change for example, or you use the --reload cli option). I need to extend this so that a specific machine can be specified. This way, when you edit settings for Machine #2 for example, only the machine #2 miner instances will reload instead of wasting cycles and time reloading all machines.  I expect to release the new changes tomorrow!

 doubt i ever be able to figure out much more than i have .  its just overwhelming to me what little bit i do with smartcoin.   finding i need to restart it dailey else i earn less not sure as to why it is .   i am gratefull to have used it even if i have to later abandon it


Jen

A big thanks to Enzomatrix for the use of his machine.

I just had my first "fully successful" test of multi-machine support (over the internet even! -- its really designed for use over a local LAN)


There is a lot to do though, as I had to manually edit a lot of the stuff the remote autodetect entered into the database, and I have some issues with output redirection over a remote connection - but still it was a 100% success!  This has turned out way more complex than I had ever imagined, I can't wait to get this to work "easily" so I can stop bashing my head against the wall every night


[EDIT]


Here is a screenshot showing smartcoin running on my machine. As you can see, I have the normal 'localhost' tab, but also an 'enzomatrix' tab!  You can also see that pretty much everything is working (minus the error bleeding through at the top of the screen).
Just thought I would share the screenshot, so users have an idea of what it will look like with multiple machines running

Just sent you a PM would love to help you in the Multi-machine cause

Does anyone have a machine that they could donate to me for a few days to test multi-machine support?  All I need is ssh access.  I'm kind of stuck at the moment as I only have 1 machine myself, so testing multi-machine support is kind of hard for me to do.  Nothing on your machine will be changed, and I can restart your local mining stuff in between bouts of experimenting.  PM me if you can help!

thanks

Thanks - I'll definitely look into poclbm support as soon as I get this multi-machine stuff committed (in a few days???)
It did work with poclbm, but looking at your output, it looks like the poclbm output has changed since I've last seen it.

Hehe ok, for some reason thought you had poclbm tested my bad.

Here is the output form the screen session launched:

Just posting some progress..

Did my first true multi-machine tests today.  It went "partially" ok so far.
A new status screen did open for the remote machine correctly.  It also showed GPU and CPU stats correctly from the remote machine.  The miner instances failed to launch, but I didn't expect that to work on the first try.  I'm going to take  a break for the rest of the day, but we're getting very close!

I understand that, but I need to see the output for myself so that I can determine what is causing <<>> do be displayed - a grep is done for a certain string in the miner instance  for the <<>> message to be displayed