Scripts to configure an Ubuntu installation to mine ZCash with EWBF Miner on an NVidia GPU, as part of the FlyPool mining pool. Remote monitoring of the miner with DataDog is supported if you have a DataDog account.
You've got to provide your own internet-connected Ubuntu installation but these scripts should be able to take care of the rest.
This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
| Ubuntu Version | Tested? | Works? |
|---|---|---|
| Desktop 17.10 | ✔ | ✔ |
| Desktop 16.04 | ||
| Server 17.10 | ✔ | ✔ |
| Server 16.04 |
- A working, contemporary Ubuntu installation (see chart above).
- It is best if this is a brand new installation with no prior configuration and no data to lose!
- A ZCash t-address
- A modern NVidia GPU
- (optional) a DataDog API key for your DataDog account
- Clone this repository. You may need to run
sudo apt-get update && apt-get install gitif the installation is brand-new. - Run
setup.sh, providing at least a ZCash t-address:./setup.sh --zcash-address tsomethingsomethingsomething- If you have a DataDog account, you can provide your API key to enable monitoring the mining rig:
./setup.sh \ --zcash-address tsomethingsomethingsomething \ --datadog-api-key f00fd00fsomethingsomething
- If you have a DataDog account, you can provide your API key to enable monitoring the mining rig:
- When setup completes successfully, run
systemctl start miner-zec-ewbf. - Done!
Note: If you change your GPU hardware or the machine's hostname, you will need to run these scripts again.
If you have a headless Ubuntu "server" installation and you want the machine to log in and start mining automatically whenever it powers on, run this command:
MINER_USER=$(whoami) envsubst < \
./resources/autologin/etc/systemd/system/[email protected]/override.conf.template | \
sudo tee /etc/systemd/system/[email protected]/override.conf \
>/dev/null
- You must not have full-disk encryption enabled.
- You must not have "home directory" encryption enabled.
- You will be unable to log out, so anyone with physical access to the machine will be able to access your account.
Each module is idempotent: You can run them multiple times without worry about messing up your system.
Installs NVidia GPU and CUDA drivers.
| Command-Line Flag | Required? | Purpose |
|---|---|---|
--nvidia-ubuntu-version / -v |
Identify the NVidia release of Ubuntu Drivers |
NVidia drivers for ubuntu are stored at http://developer.download.nvidia.com/compute/cuda/repos.
However, they are not updated for every Ubuntu release.
It may be that the latest NVidia drivers are 1704, but your Ubuntu installation is something newer, like 17.10.
In this case, use the --nvidia-ubuntu-version flag to provide the correct identifier, e.g. --nvidia-ubuntu-version 1704.
The script will instruct you to do this if necessary.
Note:
Nvidia driver installation itself is not actully idempotent; the script leaves a breadcrumb in ~/.nvidia-version
so it can know if the correct drivers have already been installed and skip installing them if they are already present.
Sets up the Ubuntu Linux environment for a primarily-unattended mining rig. Installs an SSH server and upgrades all installed packages.
Installs the EWBF CUDA ZCash Miner as a system service, connected to the FlyPool ZCash Mining Pool.
| Command-Line Flag | Required? | Purpose |
|---|---|---|
--zcash-address / -z |
✔ | The "t-address" to send earned coins to. |
--fan-during-mining / -f |
A %speed to set GPU fans to when mining starts. |
If set, the GPU fans will be set to this % of maximum possible speed when the mining service is active and automatically managed by the GPU when mining is inactive.
If un-set, the GPU fans will be automatically managed by the GPU at all times.
This uses the nvidia-settings tool which can be finicky about requiring a GUI environment.
It is recommended to not use this flag unless your GPUs' automatic fan control is very wrong.
Installs DataDog monitoring of mining activities.
| Command-Line Flag | Required? | Purpose |
|---|---|---|
--datadog-api-key / -d |
Your DataDog account's API key. |
If you provide a DataDog API key, this script will install the dd-agent and set up monitoring of the various aspects of your mining operation.
Checks:
- Config:
/etc/dd-agent/conf.d/systemd-unit.yaml - Check:
/etc/dd-agent/checks.d/systemd-unit.py - Metric:
systemd.<unit name>
| Configuration Key | Type | Purpose |
|---|---|---|
unit |
instance |
The systemd unit to check |
This can actually monitor any systemd process! By default, it monitors the miner that would be installed by the miner-installation script.
You can add any systemd units as an "instance" to monitor that one, too:
/etc/dd-agent/conf.d/systemd-unit.yaml
init_config:
instances:
- unit: miner-zec-ewbf
- unit: some-other-critical-service- Config:
/etc/dd-agent/conf.d/nvidia-gpu.yaml - Check:
/etc/dd-agent/checks.d/nvidia-gpu.py - Metric: (subset of
nvidia-smi --query-gpu)
| Configuration Key | Type | Purpose |
|---|---|---|
gpu_metric_names |
init_config |
The nvidia-smi metrics to report |
This uses the nvidia-smi tool to query the GPUs.
A comma-separated list of metrics to query is provided in the YAML file.
These are provided directly to the --query-gpu flag of nvidia-smi.
A reasonable list is provided by default.
Any metric that can be returned by --query-gpu can be added.
Metrics are reported to DataDog with the same name as used in the configuration file, and tagged with the GPU index.
If the only thing you cared about was GPU temperature, you could reduce the provided YAML to
/etc/dd-agent/conf.d/nvidia-gpu.yaml
init_config:
gpu_metric_names: "temperature.gpu"
instances:
[{}]- Config:
/etc/dd-agent/conf.d/systemd-unit-hashrate.yaml - Check:
/etc/dd-agent/checks.d/systemd-unit-hashrate.py - Metric:
gpu.hashrate
| Configuration Key | Type | Purpose |
|---|---|---|
max_log_line_age_minutes |
init_config |
Prevent stale logs from producing metrics |
unit |
instance |
The systemd unit logs to read |
gpu_hashrate_regex |
instance |
Parse hashrate from the systemd logs |
Reads the system journal of a systemd unit with journalctl to see miner output
and finds the parts where the miner reports per-gpu hash rate.
Multiple instances may be configured to check hash rate from additional systemd units, e.g.
/etc/dd-agent/conf.d/systemd-unit-hashrate.yaml
init_config:
max_log_line_age_minutes: 5
instances:
- unit: miner-zec-ewbf
gpu_hashrate_regex: 'GPU([0-9]+): ([0-9]+) Sol/s'
- unit: some-other-miner:
gpu_hashrate_regex: 'gpu #([0-9]+) - ([0-9]+) hashes'journaltcl buffers output occasionally.
DataDog only checks occasionally.
This check will read the latest entries in the journal for the miner from the bottom-up, until it finds a line that reports per-GPU hashrate. However, even if mining has stopped, these log lines will still be present.
This setting allows a maximum age to be configured to avoid the situation where mining ended long ago, but there is still a log line that shows a non-zero hash rate.
The regular expression used to find the per-GPU metrics is configurable. The first capture group should be the GPU index, and the second capture group should be the metric.
If you run these scripts off of a USB drive, that drive must use a file system that supports POSIX-style permissoins (i.e. can chmod).
If your USB drive does not (e.g. if it's formatted with the very-common fat32 format), you must copy these scripts off of it, first.
If you provided the --fan-during-mining <%speed> flag while installing
but the fans are not actually operating at that speed while mining
(as shown by nvidia-smi --query-gpu=fan.speed --format=csv), check
the logs for the miner with journalctl -u miner-zec-ewbf --lines 100.
If you see something like this:
bash[1382]: ERROR: Error assigning value 100 to attribute 'GPUTargetFanSpeed'
bash[1382]: (hostname:0[fan:0]) as specified in assignment
bash[1382]: '[fan:0]/GPUTargetFanSpeed=100' (Unknown Error).
There can be at least two different causes:
If /etc/X11/xorg.conf does not contain the "Coolbits" setting, nvidia-settings won't allow you to control fans directly.
Run nvidia-xconfig --cool-bits=4 and then check /etc/X11/xorg.conf for a section similar to this:
Section "Screen"
Identifier "Screen1"
Device "Device1"
Monitor "Monitor1"
DefaultDepth 24
Option "Coolbits" "4"
Option "AllowEmptyInitialConfiguration" "True"
SubSection "Display"
Depth 24
EndSubSection
EndSection
If it's there, then either you just fixed the issue or that wasn't the issue. You'll need to restart X before the changes take effect; the simplest way to do that is to reboot.
It could be that something's gone wrong with the graphics drivers.
This can happen if you apt-get install ... nvidia drivers more than one time.
The only way I know to fix this is to completely remove them, reboot, and reinstall:
sudo apt-get purge -y nvidia-* cuda* nouveau*
sudo apt-get autoremove -y
rm ~/.nvidia-version
sudo reboot
After that, run the installation scripts again.