djenriquez/cudamon repository overview

⁠cudamon

NVIDIA Cuda Monitor

⁠Summary

CUDAMon is an NVIDIA GPU monitor with the ability to notify on problematic events. CUDAMon is configured to monitor a metric against a threshold, then notify you incase that threshold is surpassed. Supported metrics:

• Temperature
• Utilization

⁠Dependencies:

• Docker-CE (17.02+ recommeded)
• NVIDIA-Docker⁠
• AWS Account (for access to SNS)

⁠Environment Variables:

• TIMEOUT_MINS: The length of time in minutes between notification alerts. Default 10.
• POLL_SECS: The length of time in seconds between each nvidia-smi poll. Default 60.
• GPU_UTIL_<ARCH>: The utilization threshold that the specific GPU model must stay above. Default 80.
• GPU_TEMP_<MODEL>: The temperature threshold that the specific GPU model must stay below. Default 75.
• SNS_TOPIC_ARN: The AWS SNS topic ARN to publish to. Note: SNS messages are published in raw message structure.
• AWS_ACCESS_KEY_ID: The AWS Access Key ID /w publish access to the SNS topic ${SNS_TOPIC_ARN}.
• AWS_SECRET_ACCESS_KEY: The AWS Secret Access Key associated with ${AWS_ACCESS_KEY_ID}.

⁠Run

docker run -d --name cudamon \
--runtime=nvidia \
-e TIMEOUT_MINS=10 \
-e POLL_SEC=15 \
-e GPU_UTIL_1070=90 \
-e GPU_UTIL_1080=90 \
-e GPU_TEMP_1080_TI=85 \
-e AWS_ACCESS_KEY_ID=${AWS_ACCESS_KEY_ID} \
-e AWS_SECRET_ACCESS_KEY=${AWS_SECRET_ACCESS_KEY} \
-e SNS_TOPIC_ARN=${SNS_TOPIC_ARN} \
djenriquez/cudamon

⁠Donations

• BTC - 33DyXVuy3R5jfLZRRpEQcXXAJ1Xz5rkGxE
• LTC - MUaov1JidbnpfeuQiSR3mtJhN3CN8Wj5g9
• ETH - 0xCBBC579Ac1Bc4868823fbBb2D8dDaFF93D619ceD
• DASH - Xy4cgJVAiHsrbeBB53NeQWk2iXKoWjBvJp
• ZEC - t1gYs8Zn2ZCFZWKZsTmZWd5bgXa9eD8M87K
• BCH - LOL