cavium - raeker/ARC-Wiki-Test GitHub Wiki

Advanced Research Computing : Cavium

Creating a Cavium Hadoop User Account

  1. First, ssh to cavium-rm01.arc-ts.umich.edu
  2. Then, run the following command:
    • sudo /usr/arcts/systems/scripts/create-cavium-hadoop-user.sh -u <uniqname> -g <group>
  3. Push the changes by running:
    • sudo /usr/arcts/systems/scripts/ansibleSync.sh
  4. Add the user to the Cavium MCommunity group:
    • hadoop-users-flux-thunderx

*** note ***

Cavium-rm01 is source of truth for users and groups on cavium hadoop

Cavium-rm01 is the true head node for cavium hadoop and has no root sq on /home and ssh root access to all cluster nodes

Sometimes you may need to create a group in both nyxb/flux and cavium hadoop, eg, twitter data staging for qmei via globus/fluxxfer but very rarely needed

Updating a Group on Cavium Hadoop

  1. First, ssh to cavium-rm01
  2. To add or remove users from groups respectively run sudo usermod -aG <group> <user> or sudo gpasswd -d <user> <group>
  3. Push the changes by running:
    • sudo /usr/arcts/systems/scripts/ansibleSync.sh

Check if a Queue has a Corresponding Group

  1. First, ssh to cavium-rm01

  2. Run less /etc/hadoop/conf/capacity-scheduler.xml

  3. Look for a line <!-- <queue> queue START -->

  4. In the property *.acl_submit_applications, look for a group name in the value line.

⚠️ **GitHub.com Fallback** ⚠️