 Show Posts
|
|
Pages: [1]
|
|
It does log which GPU had the low temp
|
|
|
|
|
I improved the script yet again because I got tired of viewgpu's readings being unreliable.
It now goes off your card temperatures, which seems to be very reliable (a card that is sick or dead will quickly cool down to idle temps).
So far this hasn't failed me!
|
|
|
|
|
When you run it manually, that process should be dead, it just means it completed successfully. Basically there's a watchdog that only gives the viewgpu command a few seconds to complete (otherwise a stuck viewgpu command will make the script just wait for it to finish indefinitely, but if viewgpu runs normally, when kill goes to kill the process, it will have already terminated successfully).
If you weren't able to SSH in, it's possible that the Kernel just blew up? At that point, nothing will help you, not even Linux magic keys. You'll have to power button if the kernel explodes.
If you ever see a situation where it doesn't reboot and a card is dead (and you can still actually interact with the system) run the viewgpu command and give the output back.
But yeah, a malfunctioning Linux kernel, whether it's a kernel panic or some other kernel explosion means that your only likely resolution is actually powering it off.
|
|
|
|
I had my GPUs stop spinning last night and the rig halted, it didn't reboot though and I'm running the script.
Here is what I see when manually running it, I'm using SMOS 1.3
/root/autoRebooter.sh: line 11: kill: (10583) - No such process
Hmmm, that line means that the viewgpu process likely completed just fine. Next time that happens, can you run viewgpu and paste me the output? It's possible that there is some failure output that the script doesn't check for yet. Some people have mentioned that coldreboot fails sometimes, perhaps I should just remove the /sbin/coldreboot and always just use Linux magic keys (echo s, echo b, etc). |
|
|
|
|
One last edit to clean things up. Currently it deals with any edge cases I've seen come up (coldreboot fails, viewgpu hangs, etc).
|
|
|
|
|
I edited the script in the OP to take care of some additional edge cases where viewgpu command might hang indefinitely.
|
|
|
|
Some people have mentioned that sometimes coldreboot can fail, while I've never had this problem, you could try using Linux magic keys as a failover if coldreboot fails, just sleep for 30 or so then #after the /sbin/coldreboot line
sleep 30
echo s > /proc/sysrq-trigger
sleep 10
echo b > /proc/sysrq-trigger |
|
|
|
|
I'm currently running queue:1, expiry: 30, and scan-time: 30
Are these okay settings, or should I change something to reduce reject rate?
|
|
|
|
I just got my rig setup a few days ago, and the one thing that really bugged me, especially when I was playing around with different settings, is that if a card was unstable and crashed, the only way to fix it was to coldreboot. It's possible that you could have a card that might only crash once every few days, but at that point you have to issue a coldreboot. What if you're away from computer? Think of all those precious coins you could be losing! Anyway, this script will take care of that for you. I'm using LTCrabbit's customized SMOS-Linux, but it should work for any other similar distro. First we'll need to make a script. You can use nano or vim or whatever you prefer, I'll write the tutorial using nano since if you're a Linux newbie it's probably the easiest way to go. Fire up a root terminal, then nano /root/autoRebooter.sh Paste the following contents into that file (make sure to edit your targetMinTemp accordingly!!!): #!/bin/bash
#Set your targeted minimum temp here, system will issue a cold
#reboot if a card temp falls below this number
targetMinTemp=50
i=0
(/opt/bamt/viewgpu | awk '{ print $2; }' | cut -c -2 > /tmp/viewgpu) & pid=$!
echo $pid
(sleep 10 && kill $pid)
sleep 15
array=(`cat /tmp/viewgpu`)
if [ ${#array[@]} -eq 0 ]; then
echo "`date +%m-%d-%Y` `uptime | awk -F, '{sub(".*ge ",x,$1);print $1}'` viewgpu command failed to run, rebooting" >> /home/$(grep '1000' /etc/passwd | cut -d ':' -f 1)/autoRebooter.log
/sbin/coldreboot &
sleep 30
echo s > /proc/sysrq-trigger
sleep 10
echo b > /proc/sysrq-trigger
fi
for temp in ${array[@]}; do
if [ $temp -lt $targetMinTemp ]; then
echo "`date +%m-%d-%Y` `uptime | awk -F, '{sub(".*ge ",x,$1);print $1}'` card number $i has stopped, its current temp is $temp, coldrebooting" >> /home/$(grep '1000' /etc/passwd | cut -d ':' -f 1)/autoRebooter.log
/sbin/coldreboot &
sleep 30
echo s > /proc/sysrq-trigger
sleep 10
echo b > /proc/sysrq-trigger
fi
i=$(($i+1))
done
Use ctrl+o to write the file out, then ctrl+x to exit nano. Next you'll need to make the script executable chmod a+x /root/autoRebooter.sh Lastly, we'll need to add a cronjob to periodically check in. I set it to run every hour. Add the following line to the end of crontab 0 */1 * * * /root/autoRebooter.sh ctrl+o to write it out, ctrl+x to save it. There you go, now you never have to worry about a crashed GPU bringing down your hashrate ever again! If you found this helpful, I'm currently at a whopping 2.3 LTC and 0.01 BTC would love to have a few fractions more! LTC: Lhb3yJGPL9dsUZ2tt5KrbNMm3pVmmA1fkbBTC: 1NKkGEsY5UwkzSmD63yBcJj9hkrS4YWsbXedit: I've made improvements incase viewgpu gets stuck or coldreboot fails, tested and verified to work! |
|
|
|
Why should the early adopters be rewarded with incredible wealth while I get a pittance?
But if you buy bitcoins today for $7 and sell them after three years for $100,000 isn't that enough reward for you? Sure, the value of a bitcoin after 3 years may still be $7 or even zero, but there is a also a chance that it really reaches to extremely high values. Right, but it still doesn't fix the gut reaction problem that even small children have. Split a candy bar with two kids 70/30, and the kid with the 30 isn't going to say, "Sweet, three tenths of a candy bar!" He's going to say, "Why did he get twice as much as me?!" I think at a base level, we all have that concept of "fairness" in our decision making processes, and so long as that isn't rectified, I don't see BTC gaining a level of appeal necessary for its success. |
|
|
|
|
Hey, I'm a software engineer who after seeing all the constant posts on Slashdot about bitcoins, decided to come see what all the fuss was about. After reading the FAQ, I immediately wanted to come and make a post containing the exact points that "Suggester" has made.
I would totally be excited to get involved in both mining and using BTC as currency if it weren't for the flaws he's outlined. I won't restate every point he made in his original post, but my own thoughts are as follows.
BTC will work fine once it gets enough people using it and accepting its worth (maybe a few slight modifications are needed), but the biggest problem is reaching that many users of BTC. As a potential user who is "late to the game" I see a system that hugely favors the early adopters--it just doesn't come off as "fair", and as such, I'm not too anxious to get involved. Why should the early adopters be rewarded with incredible wealth while I get a pittance?
Anyway, I'm not an expert economist, but I'm guessing my thoughts after reading and understanding how the bitcoin system works are going to mirror those of countless others who stumble upon BTC this late in the game and then investigate how it works. If you want the currency to succeed, you need to attract myriad people, and the only way I see that you're going to do that is to avoid coming off as a "ponzi" scheme, which, even if BTC isn't, it certainly comes off as one at first glance (huge rewards for early adopters, little to no incentive to adopt as time passes).
edit: I know the retort is going to be, "Bitcoin isn't about mining, it's about having a stable currency for exchange." And to that I say, "I agree!" Which is why you need to remove the huge incentive that early adopters get. With that gone, I wouldn't even want to mine, I would just want to immediately get involved in utilizing the currency for its intended purpose.
It isn't so much that the system is flawed, but that from a psychological perspective, people have a real problem getting involved in a currency system that looks like a ponzi scheme--my gut reaction is to reject a system that provides substantial wealth to early adopters, while providing significantly less incentive for new adopters to get on board.
|
|
|
|
|
