search

microsoft / pai

Resource scheduling and cluster management for AI
https://openpai.readthedocs.io
MIT License
2.64k stars 548 forks source link

Does the master-machine must have more than 40Gi memory size？ #5610

Open JohanOu opened 3 years ago

JohanOu commented 3 years ago

Short summary about the issue/question: I got a error like this when I install kubernetes. fatal: [pai-master]: FAILED! => {"changed": false, "msg": "The following requirements are not met: OpenPAI's infra node should have 40 Gi free memory for service"} I set 10Gi in layout.yaml because my master-machine just have 16G memory size. Can I install openpai?How to install? Thanks very much.

OpenPAI Environment:

Anything else we need to know:

siaimes commented 3 years ago 
Please press ENTER to stop the script, check the log, and modify the cluster setting to meet the requirements.
If you are very sure about the configuration, and still want to continue, you can type in "continue" to force the script to proceed.

You can just input continue to ignore this fatal and continue the install.

JohanOu commented 3 years ago 
Please press ENTER to stop the script, check the log, and modify the cluster setting to meet the requirements.
If you are very sure about the configuration, and still want to continue, you can type in "continue" to force the script to proceed.

You can just input continue to ignore this fatal and continue the install.

Thanks!I meet a new question: Are your cluster is in Azure cloud or not? (Y/N) (case sensitive)N cp: cannot create regular file '/home/pcl/pai-deploy/kubespray/inventory/pai/': No such file or directory The pai-deploy just have a directory:cluster-cfg how to solve it?

siaimes commented 3 years ago

Maybe your network cannot access github.com, which causes the clone project kubespray to be unsuccessful.

You can run these few lines of code yourself to see what the result is. https://github.com/microsoft/pai/blob/815ba939e6ee41192da432d1f54194da10b6fde1/contrib/kubespray/script/environment.sh#L18-L20