tytanick, takasugi, reelen, Can help to do a short ssh primer on how to use the atiflash in smOS to be shared amongst miners here. Because many of us were ex-Windows users with not much Linux exposure. Eg. How to upload ROM file in ssh? In future, making a wizard-like atiflash process in smOS would be nice if Tytanick plan to do it this way. Thanks
This tutorial would be nice in PDF, but i really dont have any cash to pay you fot this. I can only load your balance in SMOS when fee service comes. tytanick, Do you think its good idea to create a library of known user submitted rom mods for smOS users to select (just like selecting miner apps). Obviously this will require strict evaluation so that smOS dont brick any GPUs. Sure, this is a potential recipe for disaster, but this smOS feature will surely take your product to the next level. Hopefully, we will have a collection of highly optimised rom mods specifically for certain coins and power consumption preferences. To have original roms by GPU model and vendor will be great too, so that we can revert GPUs back to original roms. Thanks
I thought about it. And it would be very nice but this generates many problems to solve. Thats why i am not ready for doing that. I can only create library and then share it. Anyway i wont download users roms without permision and also i think that users wont share this for free. We can talk about this feature much later, i still need more things to do first in SMOS. I want to implement private key signing of more crucial actions like flashing roms or applying updates. This would make sure that even if my server would be hacked, then hacker wont do anything special becasue it wont have private key. This feature is most important for me. Also becasue of this i would need to secure root account on rigs and not give password for users or create possibility to apply own private key to your rigs. FLASHING TUTORIAL To flash its important to do following: - ssh as user "miner" and execute: screen -x miner .... and then press ctrl+c few times until whole screen exits. - go to root writing sudo su - execute command: dmesg -c - this will clear system log - this is where my syftoware watchdog is looking for some critical erros, if it finds them it execute reboot command. - go to atiflash directory and atiflash -p 0 custom_bios.rom atiflash -p 1 custom_bios.rom etcI am successfully flashing ROMs on SImplemining but my way is a bit more involved (but it sure is prettier also) -- I installed a desktop to run next to the Terminal environment. I installed ATIflash and Teamviewer on all the rigs and can log in remotely and stop the mining by switching to the Terminal View and then log back into the Desktop. This is using the Rx Version - I had to update the drivers on it for 2 reasons: 1- I didn't want 2 versions for different cards 2- Because neither the R or the Rx version worked well with Pro Duo's. In order to do that easily, I changed the original Password for Miner then installed a Gnome Desktop, then installed amdpro-gpu drivers. Since I had already installed a desktop I figured I might as well install something to make it more useful hence the TeamViewer and AtiFlash
Its makes completly no sense installing more packages that uses cpu, ram, and makes booting longer. You have ssh for this Also be adviced that my auto updates can easly destroy what you installed. I dont advice to do own modifications and i cant support them as it would be very difficult to mantain all different versions of my system. Also. I have possibility to buy tool for 0.25 BTC which can overclock gpus by setting specific core and memory clocks in MHz. Still i dont have cash to buy this tool. I gues it would be nice to have possibility to set specific clocks and not dpm levels right ? Propably need to impelemnt those fees fast and gather some cash to buy this tool
|
|
|
UPDATE v1074: - atiflash for linux - for now its only program located at /root/utils/atiflash If you know putty (ssh) then you can login into rig and flash custom bios remotly. The flashing from dashboard will be propably implemented in future but not soon. Right now you can do the flashing only by direct ssh into rigs.
|
|
|
T - can you consider the following features:
1. counter/countdown displayed for circles/rigs that has begun reset/power-cycle process. This way, at dashboard we can see what process has begun and have taken place. This is key for troubleshooting. WIthout this, I dont even know if SRR is working or running -- and assume its working. Over at smOS web dashboard, it would have been nice to have the power/reset control available but as explained by you -- smOS don't communicate with SRR.
2. for me, when SRR need to be activated on a rig that means miner has failed to restart the miner on its own, meaning problem may be related to other than miner, eg. memory leaks/corruption, network issues etc. Instead of reset, I would almost certainly prefer power cycle -- meaning power down and x minutes late power up.
3. When running Claymore miners, I also have his monitoring utility running in the background. This way I keep tabs with whats going on at miner app level. Claymore can also do run bat file when miner hangs due to oppencl errors - not sure how you plan to use this current utility to somewhat compatible with SRR for Claymore miners only. Maybe this is not necessary but his utility has many key info like how many times the miner resetted, how many time pool failovered etc.
4. I have re-arranged my SRR units and slot numbering again so that I didnt miss anything - so that I can correctly activate all SRRs to value more than 0 for wd delay; but I think I really need the "2500 secs" watchdog delay because my non-RX cards are STILL being triggered to reset unnecessarily (or too early). I cant figure out why this situation doesnt effect my RX480 rigs -- they respond well to SRR even at 250 secs. I can only conclude this may be how my network environment is laid out. So looking forward to the v2 fo SRR and enhancements.
1. See what i can do, its good feature. About controlling SRR i will be getting into it, i have 2 concepts of doing that and i will implement one or even two of them. a) sending commands that for example to shutdown rig number 1 but sending this command thru another working rig. b) it is propably possible to make SRR comunicating with my SimpleMiningOS dashboard, i will look into this mater and see if i can implement it but dont know ETA. 2. In my case, like 80% problems are solved by fast restart, and only some like 20% need cold reset (shutdown, wait 5 minutes, power on) I will check if i can implement some kind of feature to preset this by own needs. 3. If gpu error occurs AND you have set (-w1 -r1 in command) the it will run reboot.sh in which there are special commands that will FORCE REBOOT your rig, and this is working pretty well most cases. In some cases if this will freeze rig, then now SRR kicks in by not getting keep allive messages. So in other words, software on rig side can solve most of problems and doest that, but if it fails and the rigs wont send keepallive message within specified number of seconds then in second line SRR starts rebooting. I think that we dont need to make it compatible as every of those features works on different level (first software reboot, and if it wont help then SRR hardware fast reboot and if it 4 tome fails then hardware cold reboot) Isnt that best idea ? I will try to make SRR restarts counter, but i will also do the restarts counter in SMOS dashboard which will be looking at rig uptime, if in next report rig will show less uptime than last one - it means reboot. I was thinking about it and i will do this. 4. 2500 seconds will be in this next release. already doing that. Also i might know why R OS is booting longer. Its booting Graphical enviroment and THEN it starts running SRR agent. It takes lot more time. in RX OS there is no graphical interface so the booting process is much faster. I thing i can speed up this agent script under R OS so you wont have this issue. thx for reporting.
|
|
|
Hello!
Two questions:
1) Two of our rigs were stuck because watchdog detected a problem with one GPUs and needed to restart the mining process, but froze there. Can we use -r 1 to force the rig to restart instead? This way, they will restart the mining process properly. Even on windows, we had that issues, -r 1 would help. I am wondering if the reboot.sh is present in the claymore folder.
2) Would it be possible to have atiflash installed on the image? (maybe it is and I didnt find it) It would be greatly appreciated and it would save us time to flash our cards instead of putting a windows drive in order to do the flashing.
Thanks!!
1. Yes it will help -w1 -r1 2. I dont have it now, but i am planning to add this. As now i can add this program to root directory and then you can do the rest via ssh. But option for flashing from dashboard will be propably possible but later on.
|
|
|
There will be very soon New version of SRR Tool + Firmware Things that will be changed:- instead of RED/GREEN circles there will be LIST also RED/GREEN with possibility to write custom names near each rig. Those info will be saved in SRR Tool directory. (wont be on device itself) - max watchdog delay increased from 250 seconds to 2500 seconds (some windows rigs on HDD needs more time to boot up and 250 isnt enough for them) - access to SRR via website (more and more functions will be with every release) - for now most crucial like on/off/reboot/long/reboot and status view. Logic of auto rebooting rigs: Curently: after timeout of "watchhdog delay" SRR will do fast reboot, again wait specified number of seconds in "watchdog delay" and so on...... New Logic: First 4 times it will fast reset rig, then if it still wont receive any keepallive packets it will be doing infinite number of times cold reset ( also name changed from "long reset" to "cold reset" as it is more obvious.) The question is:What is most desired behaviour for auto rebooting rigs ? 4 times fast reset and then infinite long resets is good choice ? Ofcourse if after 2nd reset rig will be allive then in next failure it will start from the beginning: 4 fast resets and infinite cold resets. Or some other configuration ? If you have other idea then tell me why you think that ? Of confirm that you think its ok ? Also if you have other requests and ideas and i forgot to write them down, please tell me or remind me
|
|
|
Hey tytanick, do you plan to support nvidia gpu's anytime soon?
Nope, i will start thinking about that not sooner than in 4 months. Setting DAG epoch #109 for GPU #1 Create GPU buffer for GPU #1 OpenCL error - cannot allocate big buffer for DAG. Check readme.txt for possible solutions. Setting DAG epoch #109 for GPU #0 Create GPU buffer for GPU #0 OpenCL error - cannot allocate big buffer for DAG. Check readme.txt for possible solutions. Setting DAG epoch #109 for GPU #2 Create GPU buffer for GPU #2 OpenCL error - cannot allocate big buffer for DAG. Check readme.txt for possible solutions. ETH: 03/05/17-08:32:13 - New job from pool.bitclouder.co:3333 ETH - Total Speed: 0.000 Mh/s, Total Shares: 0, Rejected: 0, Time: 00:00 ETH: GPU0 0.000 Mh/s, GPU1 0.000 Mh/s, GPU2 0.000 Mh/s, GPU3 0.000 Mh/s, GPU4 0.000 Mh/s Setting DAG epoch #109 for GPU #4 GPU 4, OpenCL error -38 - cannot write buffer for DAG Setting DAG epoch #109 for GPU #3 GPU 3, OpenCL error -38 - cannot write buffer for DAG Setting DAG epoch #109 for GPU #0 GPU 0, OpenCL error -38 - cannot write buffer for DAG Setting DAG epoch #109 for GPU #1 GPU 1, OpenCL error -38 - cannot write buffer for DAG Setting DAG epoch #109 for GPU #2 GPU 2, OpenCL error -38 - cannot write buffer for DAG GPU0 t=17C fan=67%, GPU1 t=18C fan=67%, GPU2 t=18C fan=67%, GPU3 t=30C fan=67%, GPU4 t=20C fan=67% WATCHDOG: GPU error, you need to restart miner Binary file (standard input) matches Close 2 GB memory in gpus ?
|
|
|
Update v1073 - claymore-zec v12.2 fix memory leak note. it will start using this after reload
|
|
|
Tytus, any plans on using the CPUs in the rigs as an option for mining? Hard thing to code or easy to do? Since at most times it makes no sense to mine the same coin / algo as your GPUs.. maybe with a seperate group in the GUI?
Cpu mining will be only as an option to mine as fee for me. There wont be mining for cpus as there is no sense of doing that.
|
|
|
Hey,
We are currently encountering a problem with the initial setup with our rigs. We can see the config.txt file and we did change the email inside and save.
After that we put it on the rig, it starts mining on your wallet like you said, but it's staying there forever after.
We spent hours to reformat, tried different ways, we always edit the config.txt with our email address and it's still mining to your address. Normally, we can see the added rig after few minutes, but it's also not showing up in the interface at all.
What else should we do?
Thanks
If above answer wont help, please write to me at admin@simplemining.netI need to check it. Also you are saying that few rigs are already working and mining to your wallet but only some are not ? I can see whats going on remotly, just need access to windows tema viewer that is on the same network as those rigs and i need rigs ip. Write email , not here ofcourse
|
|
|
After full 24++ hours -- my farm is all green with SRR!
I know of 1 or 2 problematic rigs that would turn red after a day or two -- and I could see that SRR dealt with it!
I have 33 rigs and 30 are connected to 4 units of SRR in a single network environment.
My settings : 0 seconds / disabled for watchdog max delay 3mins long reset. Save and write the above info to SRR. Set the rig details in smOS SRR menu.
After further clarifications from Tytanick -- I now understand how this gadget works.
Very simple solution made into a product. Easier said than done though.
Glad that Tytanick made the effort to make this product a reality.
More improvements to come - I hope the features discussed in the thread will be in next release
Hi. You can check which rigs rebooted by going to dashboard and fovering over "ONLINE" - you will see uptime. Also if you have 0 seconds max delay in resetter then its disabled. So if you want to use it then you need to writle like 250 here ? You know that right ?
|
|
|
its too complicated for me to do undervolting tool I would need low level programmer or perhaps i can buy this tool from someone but i suppose it will cost too much if even possible to buy. If there would be public tool i would implement it for sure but there is not About auto updates. Auto updated cant be stopped, i need them to mantain consisten system. If i will implement new feature and update server but your rigs wont update then reading stats or chaning settings wont work. I can minimalize this in future and make betatest updates like with the kernel one. Anyway i cant do this right now, it would create too much work for me. But be sure that if i want release production update then i am testing it first on my development rig. So right now i wont stop auto-updates but in future i will try to figure something out. Anyway thing i wont do: - reload or reboot someones rig without permision. - so i wont reload miner if new miner version comes out - you need to do it yourself and check if new miner version doesnt cause any problems The system with non-auto updates is ethOS. My idea is to make all rigs consistent and up to date. I am nod bad guy and i wont make your rigs zombiee or mining for me - that you can be certain.
|
|
|
Tytanick....
on the smOS, is it possible to show SRR activities.
For example:
When SRR is triggered, maybe that row for that a rig, has some indicator/flashing to show that SRR in is progress.
I have my dashboard displayed 24x7 for monitoring purposes and it will be good to see SRR is working.
Thanks
That would be difficult to do on SM OS as SRR isnt communicating with it. To test if its working its best to unplug rig 1, and see if it reboots after specified time, then rig 2 etc etc Also there wound need to be hard hang of rig becasue if only gpu hangs and system sees that it triggers software linux reboot. Anyway try to uplugging RJ45 from the rig you want to check if it will reset. Can the SRR turn on a rig from a cold start automatically? Right now not. Still thinkg what is multi best option. Now its normal resetting in loop.
|
|
|
Tytanick....
on the smOS, is it possible to show SRR activities.
For example:
When SRR is triggered, maybe that row for that a rig, has some indicator/flashing to show that SRR in is progress.
I have my dashboard displayed 24x7 for monitoring purposes and it will be good to see SRR is working.
Thanks
That would be difficult to do on SM OS as SRR isnt communicating with it. To test if its working its best to unplug rig 1, and see if it reboots after specified time, then rig 2 etc etc Also there wound need to be hard hang of rig becasue if only gpu hangs and system sees that it triggers software linux reboot. Anyway try to uplugging RJ45 from the rig you want to check if it will reset.
|
|
|
Any chance of an API or letting us know some "hack" to control the rigs through some script we have running (other than killing SRR-Agent-Linux-v2.sh on each machine)? I can quickly see where I will be monitoring my rigs at a pool and will need to reboot rigs based on what a script sees at the pool. The GUI is nice for lots of people, but some way to talk to the SRR through code for me takes it to the next level! Any chance of an API or simply some hints at how to hack talking to the SRR? For example, someway to talk to the SRR and tell it to reboot machine #5 would be ideal. Also, what is the login/password for the SRR? If I can get on the SRR I am sure I could hack something together. Overall, it looks to be a great product! Yes, sure. You can hack. There is already API which is using SRR Tool for communicating and local agent scripts. Ok so here is reset script: #!/bin/bash
# REQUIRED packages: socat
# INPUTY serial="000002" port="3"
serial=`echo $serial | xargs` port=`echo $port | xargs` port=`printf %02X $(( ${port} -1 ))`
firstByte="FF" byteCount="0008" action="53" mac="485053$serial"
checksum=`printf %02X $(( (0x${byteCount:0:2} + 0x${byteCount:2:2} + 0x$action + 0x${mac:0:2} + 0x${mac:2:2} + 0x${mac:4:2} + 0x${mac:6:2} + 0x${mac:8:2} + 0x${mac:10:2} + 0x$port)%0x100 ))` packet="$firstByte$byteCount$action$mac$port$checksum" echo "Wysyłam pakiet o następującej zawartości: $packet"
echo -n "$packet" | xxd -r -p |socat - UDP-DATAGRAM:255.255.255.255:1051,broadcast If you want to turn ON rig then change action to 51 If you want to turn OFF rig then change action to 52 Here is list you can easly use with this script 0x51 - turn on rig 0x52 - turn off rig 0x53 - fast reset rig 0x55 - keepalive 0x58 - long reset Rest is little bit more difficult like reading and writing settings - for this better use Windows Tool
|
|
|
Voltage Offset is a special value that mostly needs to be added to custom mods. Offset is basically value that will be undervolted from default values. So for example offset -125mv will substract -125mv from all dpm levels in core level. This is little difficult but i have some guys that are doing this for me. Pascal is in deed profitable in dual mining. Last time i checked it took like 10-15% more wattage. It didnt lowered any eth mining speed. And by mining pasc profit was like 20% bigger. IS still is. check this for example by using my command claymore-eth miner and this command:
-wd 1 -r 1 -epool stratum+tcp://79.98.144.28:8008 -ewal 0xa35b062b4d386e148dba1417e16b82d9ca1d360d -esm 0 -epsw x -allpools 1 -mport -3333 -dpool stratum+tcp://pasc-eu1.nanopool.org:15555 -dwal 86646-64.442a5627fbc295c2.$rigName/tytanick@gmail.com -dpsw x -dcoin pasc -dcri 15 -allpools 1
Yep, normally without undervolt offset 6x470 would take like 1000 wats, mine with custom bios mod are taking about 750 watts with ETH only. But with ETH+PASC its more 10-15% but it is different thing. Anyway, offset is the thing thatforces card to take less watts under linux. Normal undervolting (changing values in polarisbios editor) doesnt work under linux. It works only under windows.
|
|
|
as for the max 250 time you have a problem but at min 0 time no problems
test 240 see if problems like that of 250 show up
test 90 see if problems show up
test 5 see if problems show up
I spoke with my SRR developer, there will be max 2500 seconds in watchdog in next Firmware release. We have limit of 255 value, so we need to to it *10. So the setting will be 10,20,30,40 seconds....... 2500seconds When time comes and i will release first firmware upgrade, i will write special instructions with changelog and what you need to know before and after upgrade.
|
|
|
Well we tried different OS on the market and we are so far impressed by your work. WindowsOS is very difficult to install and manage at large scale, so we also tried ethOS at some point, but it was a total disaster for us in stability and performance. We were kinda stuck on Windows since that time and scared of any linux OS!
We might try to deploy on more rigs when we get the chance and post the results here!
@Citronick Yes you are right about these inputs. I will def give optiminer a shot in linux. Thanks:)
About RX OS version, Dod you do beta kernel upgrade ? If not, do it to have even more Hashrate but i guess you already did it ? ethOS is much longer on market, like 2 years ? While i started like 7 months ago Is ethOS disaster ? I bought one license to test it out but never had time to do so so i cant tell. Anyway from where i can see my OS is focusing on SIMPLE while they have multiple configurations and much harder to manage. OK, test my OS and give feedback. After this grouping feature i will be also implementing very slowly some kind of fee service - its time to harvest No, seriously, it will be very slow process with no obligations and many free periods etc etc... Personally i am using eth+zec with 15 dcri Also about power consumption. Its worth to mention that in linux undervolting in tables (like editing them in polarisbios editor WONT WORK) Only way to have less voltage is by doing voltage offset. I personally am using them and i have RX470 4GB with 27MH/s each GPU, and 780 watt from wall from whole 6 gpu rig.
|
|
|
RED are my reply Tytanick - I have finally finished installing the SRRs and have some questions and assumptions below: can you help go through and see if I did the correct steps. My setup is multiple SRR in one network/farm. Thanks! 0. All of my rigs are on smOS. And I use the Windows SRR Tool to manage the SRRs. 1. If I am not using any Windows rigs to be connected to SRR, do I need to do anything in the "LOCAL AGENT" part in SRR Tool? Nope, Local agent part is only for Windows rigs1. What is "PC Number" on Windows SRR Tool? This is slot number in SRR 1,2,3,4,5,6,7,8 (you have printed on board jp1 jp2....), This misleading test will be fixed in next SRR release2. If I have multiple SRRs in the same network -- do I change the UDP port for each SRR? No, dont change this 3. My guess is that for watchdog to work, you have to have the Windows SRR Tool to be opened.... or watchdog is running in service background and will work even with Windows SRR Tool NOT launched? You can close SRR Tool, its only for configuring SRR Board, and then after you write config you can close it. Watchdog service is running inside SRR Board.4. I have renamed the SRR hostname according to my environment, and also changed the IP address to 10.88.88.1, 10.88.88.2 and so on. On, its up to you how to name them Also IP address is optional, it will be more used when Web panel for SRR will come with future firmware release. As now all communication happens via broadcast (it uses MAC to communicate). IP here is optional.5. I am using 1 single PC to launch 4 SRR Tool windows (because I am using 4 x SRRs on my farm)... on first SRR, I have installed the SRR agent in LOCAL AGENT. Pressing SAVE will enable WATCHDOG and "CONFIGURED with :00000x". Do I need to do the same process for the other SRRs? It doesnt seem to give me any option to install in setting up the 2nd SRR. Local Agent is still only for WINDOWS RIGS. As you dont have them you should not use this at all. Anyway if you are using windows rigs then Open SRR Tool on windows rig1 and set SERIAL number and port number of SRR that this rig is connected to. As you are using my Simplemining OS, just click near each rig MORE -> SRR. If your rig is connected to SRR with serial 000034 to port number 5 then enter in those options serial 000034 and slot 56. After installing and assigning the SRR details in smOS, I could fully control all rigs. Fast, long resets and powerdown, power on - no issues. If I choose watchdog delay = 0, the farm is stable and responding well to manual actions. However, when I put 250 seconds in watchdog delay, I can see several rigs are randomly going offline and online - I am assuming watchdog is working.... but seems strange that this happens whenever I choose 250 seconds max delay. For now I use watchdog delay =0 and it doesnt do anything .... and I am assuming watchdog is listening to all the rigs attached to the SRR unit. Hmm strange. It will reset only if there is a problemwith specific rig. If your rigs are resetting randomly with no apparent reason then something is wrong. Please check if you have good options written in simpelmining dashboard config. So is you are using SRR 000007 with port 5, check if resetting this rig manually will work and it resets 5th and not other rig. If everything is set right then it wont reset rig until something is wrong with it. Please check if port number you entered in config corresponds with port number that is written on SRR Board itself near this slot. Also all rigs and SRR are in the same network right ? If all those things are correct and still you have problem then please write to me at admin@simplemining.net and i will check all this via team viewer Meaby bug, meaby misconfigured, i need to check it. 7.
|
|
|
|