Skip to content

Latest commit

 

History

History
131 lines (82 loc) · 5.5 KB

faq.md

File metadata and controls

131 lines (82 loc) · 5.5 KB
Raw

OpenPAI FAQs

Table of Contents

  1. User job related FAQs
  2. Deploy and maintenance related FAQs

User job related FAQs

Q: If user find a job to retry multiple times, how to diagnose the cause

A: Users can find historical job logs through yarn. Please check issue-1072's answer and job log doc's section:Diagnostic job retried many times reason introduction.

Q: How to diagnose job problems

A: Please check troubleshooting_job.md's introduction.

Q: How to use private docker registry job image when submitting an OpenPAI job

A: Please refer job_tutorial.md to config the auth file at job submit json file:

If you're using a private Docker registry which needs authentication for image pull and is different from the registry used during deployment, please create an authentication file in the following format, upload it to HDFS and specify the path in authFile parameter in config file.

  • (1) Create an authFile

authfile content:

userprivateimage.azurecr.io
username
password

Note: userprivateimage.azurecr.io is docker_registry_server

File path at hdfs example: hdfs://master_ip:9000/user/paidemo/authfile

  • (3) Specify the path in authFile paramete

OpenPAI job json file example:

{
  "jobName": "paidemo",
  "image": "userprivateimage.azurecr.io/demo4pai:test",
  "dataDir": "hdfs://master_ip:9000/user/paidemo/data",
  "outputDir": "hdfs://master_ip:9000/user/paidemo/output",
  "codeDir": "hdfs://master_ip:9000/user/paidemo/code",
  "authFile":"hdfs://master_ip:9000/user/paidemo/authfile",
  "taskRoles": [
    {
      "name": "demo4pai",
      "taskNumber": 1,
      "cpuNumber": 2,
      "memoryMB": 8192,
      "gpuNumber": 1,
      "command": " cd /home/test && bash train.sh"
    }
  ]
}

NOTE:

  • If you're using a private registry at Docker Hub, you should use docker.io for docker_registry_server field in the authentication file.
  • Related issue: 1125

Q: How many jobs does PAI support?

A: According to the default configuration, PAI supports 60k jobs, including waiting/running/finished jobs.

The limitation is because we only reserve so much memory resource for PAI services, it may be enough in typical scenario.

For example, user may have hundreds jobs running, thousands jobs waiting, and tens of thousands jobs finished.

Q: How to solve failed call to cuInit: CUresult(-1)

You should check LD_LIBRARY_PATH in your job container by using export command. It should have one of following:

  • /usr/local/nvidia/lib
  • /usr/local/nvidia/lib64
  • /usr/local/cuda/extras/CUPTI/lib
  • /usr/local/cuda/extras/CUPTI/lib64

You can add path to LD_LIBRARY_PATH in your Dockerfile like:

ENV LD_LIBRARY_PATH=/usr/local/nvidia/lib:/usr/local/nvidia/lib64:/usr/local/cuda/extras/CUPTI/lib:/usr/local/cuda/extras/CUPTI/lib64:$LD_LIBRARY_PATH

If probelm remains, you can try this script to self diagnose the problem.

Deploy and maintenance related FAQs

Q: Why not recommend deploying the master node to the GPU server and running the job?

A: It is not recommended to run the job on the master node in order to avoid overload on the master node and affect the stability of the cluster.

Q: When OpenPAI has multiple master nodes, can the master node be deployed on multiple subnets, and they can still access normally?

A: We recommend deploying them on the same subnet. In theory, as long as the network is interoperable, it can be deployed. Considering the high communication requirements of the cluster, the network delay of different subnets is usually high, and the network is often inaccessible.

Q: To improve the cluster usage, user would like to see a VC can use up all cluster resource if others don’t use it.

A: By default, a VC can use up all cluster resource if others don’t use it. OpenPAI use capacity scheduler of YARN for resource allocation. maximum-capacity defines a limit beyond which a queue cannot use the capacity of the cluster. This provides a means to limit how much excess capacity a queue can use. Default value of -1 implies a queue can use complete capacity of the cluster.

Q: How to configure virtual cluster capacity?

A: By webportal

Q: What should the admin do when get NodeFilesystemUsage firing?

A: NodeFilesystemUsage firing is caused by disk pressure. The root cause may be HDFS, docker data cache or user data cache. when this firing happens, the admin should do the following steps mitigate the data load.

  • (1) Check HDFS.

Check Hadoop Datanodes (url example: http:// hadoop namenode/dfshealth.html#tab-datanode). If only a few nodes are overused, run "hdfs balancer -policy datanode" to rebalance the data among data nodes via HDFS balancer. If most nodes are in heavy usage, manually clean some data before do rebalance.

  • (2) Check docker cached data.

Run "docker system df" to see how much space can be reclaimed. If the space is large, run command "docker system prune -a" to clean the cache.

  • (3) Check /tmp directory.

Run command "du -h / | awk '$1~/[0-9]*G/{print $0}'" to list all the directory which consumes more than 1G spaces. Clean the data not used any more.