problem in implementing BYOC in multi container under one end point

Question

I am facing issue in implementing BYOC with multi container with one end point. I am training my own model in Container and when I am running this container singularly it is working but integrating with multi container it is giving issues in endpoint creation "ping health check not passed"

Answer

If you're using a modest instance type to host multiple models, perhaps you might be struggling with CPU/memory resource contention [as mentioned here in the docs](https://docs.aws.amazon.com/sagemaker/latest/dg/multi-container-direct.html#multi-container-troubleshooting)? SageMaker recently [launched configurable health-check time-outs](https://aws.amazon.com/about-aws/whats-new/2022/09/amazon-sagemaker-deploying-large-models-volume-size-timeout-quotas/), which may be interesting to you if your models or containers are particularly large / slow to set up (see the [CreateEndpointConfig API](https://docs.aws.amazon.com/sagemaker/latest/APIReference/API_ProductionVariant.html#sagemaker-Type-ProductionVariant-ContainerStartupHealthCheckTimeoutInSeconds))... But I'm not certain whether this configuration is supported for Multi-Container Endpoints at the moment.

problem in implementing BYOC in multi container under one end point

Contenus pertinents