Install ThoughtSpot using Ansible on clusters that use RHEL

This is an alternative installation method to the Online cluster RHEL installation and Offline cluster RHEL installation. If your organization requires that privilege escalation take place through an external tool that integrates with Ansible, follow the steps in this article.

Before starting the installation, complete the prerequisites.

Prepare system settings

There are certain system settings, specific to this type of RHEL installation, that you must configure. Configure these settings after completing the prerequisites.

  1. Install the required yum, Python 3, and R packages on each node of the ThoughtSpot cluster: Packages installed with ThoughtSpot for RHEL and OEL.

  2. If you are using a physical machine or appliance, you must disable the disk write cache. On each ThoughtSpot-specific mounted drive, run the following command:

    hdparm -W 0 /dev/sdX

  3. Configure journalctl log rotation:

    1. Create a /var/log/journal directory.

    2. In that directory, run systemd-tempfiles –create –prefix /var/log/journal.

    3. Modify the journald.conf file to specify Storage=Persistent under the [Journal] section.

    4. Set up regular journal log rotation.

  4. If you install the nfs-utils yum package, ensure that the rpc.statd port is bound to a fixed port to prevent conflicts. You can check this in /etc/sysconfig/nfs. Both STATD_PORT and STATD_OUTGOING_PORT should equal a fixed number.

  5. Set both net.core.rmem_max and net.core.wmem_max to 3145728. To check this setting, run the following commands:

    sysctl net.core.wmem_max
sysctl net.core.rmem_max

    To ensure this remains true even after reboots, set these values in a file under /etc/sysctl.d.

  6. Ensure that the fs.file-max sysctl setting is the default for linux, and that the user.max_user_namespaces is set to 5000 or higher.

    To ensure this remains true even after reboots, set these values in a file under /etc/sysctl.d.

  7. Set up the kernel cores sysctl setting to create dump files on crashes.

    To ensure this remains true even after reboots, set these values in a file under /etc/sysctl.d.

  8. Set the soft and hard nofile ulimit values to 524288 or higher. ThoughtSpot prefers the soft core to be unlimited. Pending signals (-i) must be 2062971 or higher.

  9. sshd config must contain AllowUsers: tspot. It must have a MaxSessions value greater than or equal to 10. Each node must be able to ssh without a password, via keys, to every other node in the cluster as the tspot user.

  10. Set the postfix settings to send email from the cluster.

  11. Configure passwordless ssh for ts_service_user.

Set up nodes

  1. Unzip the ansible tarball provided by ThoughtSpot Support. The ansible tarball contains the following yaml files:

    • ts-cluster-create-A.yaml: Untars the release, sets up the environment, and copies the release toolchain to all nodes.

    • ts-cluster-create-full-flow.yaml: Triggers the cluster installation. Calls ts-cluster-create-A.yaml, hadoop-ansible/install.yaml, ts-setup.yaml, and ts-no-sudo-cluster-install.yaml. After the release tarball is installed, it calls installer_main from the coordinator node. The installer_main pushes all the services and finishes the installation.

    • ts-no-sudo-cluster-install.yaml: Installs the release on all nodes. Starts up ThoughtSpot core services on all nodes.

    • ts-packages.yaml: Checks that postgres was installed correctly and sets up atop.

    • ts-setup.yaml: Sets up the system settings, including nginx, log rotation, and environment settings. Ensures python3 packages are installed. Calls ts-systemd.yaml and ts-packages.yaml.

    • ts-systemd.yaml: Sets up systemd services.

    • hadoop-ansible/install.yaml: Installs Zookeeper and HDFS on the cluster.

  2. Ensure that the nodes are correctly initialized. Run the following command twice from ts_service_id, once as a root user and once as an admin user. The output of this command informs you of any missing configurations and provides recommended actions.

    ./os_settings.sh

  3. Configure passwordless ssh for ts_service_user.. Set up the ThoughtSpot directories that require root privileges. Run the following commands:

    mkdir -p /export/scaligent /export/thoughtspot
ln -s /export/scaligent /usr/local/scaligent
ln -s /export/thoughtspot /etc/thoughtspot
chown -h $ts_service_id:$ts_group_id /export /export/scaligent /usr/local/scaligent /etc/thoughtspot /export/thoughtspot

  4. Fix the postgresql symlinks after installing it.

    mkdir -p /usr/lib/postgresql
ln -s /usr/pgsql-11 /usr/lib/postgresql/11

  5. Set up symlink for localtime.

    ln -s /etc/localtime /usr/share/zoneinfo/localtime

Create the cluster

Cluster creation takes about 60 to 90 minutes.

The node that acts as the coordinator node must be part of the cluster. It is defined by the var: local_node in the Ansible vars. The release tarball <software_release.tar.gz> must be present in the coordinator node and must be specified by the tarball_location var in the coordinator node.

  1. Untar the ansible.tar.gz file on the node from which the ansible playbook will be run, if you have not already. If you are using the ansible tower for running the playbook, this should be the ansible tower machine.

  2. Rename cluster_create_hosts.sample to cluster_create_hosts.yaml.

  3. Update the Ansible vars in cluster_create_hosts.yaml with your own specific values. See Sample Ansible vars.

  4. Run the ansible-playbook. This installs ThoughtSpot software on all nodes. Run the following command on any cluster node:

    ansible-playbook -i cluster_create_hosts.yaml ts-cluster-create-full-flow.yaml

    If privilege escalation requires a password, add the -K option to the ansible-playbook command.

  5. If there is a failure at any point, fix the failure and rerun the ansible-playbook command. Bypass the completed steps using the --start-at-task option, specifying the task at which to start.

Sample Ansible vars

This is the definition of the cluster_create_hosts.sample file that is present in your Ansible tarball. You must replace anything in this file within <> with your own specific information.

all:
  hosts:
    # List of IPs of the nodes in the cluster
    <node_ip1>:
    <node_ip2>:
  vars:
    ssh_user: <ts_service_user>
    username: <ts_service_user>
    groupname: <ts_service_group>
    env: {}
    ssh_private_key: <private key use for ssh>
    tarball_location: <Release tarball complete path>
    cluster_id: <Cluster id. Provided by ThoughtSpot.>
    cluster_name: <Cluster name>
    ramdisk_size: <size of ramdisk for falcon> # The default is 50619136k (50Gb)
    # ThoughtSpot variables. Do not modify.
    release_location: /export/releases/root/
    installer_spec_path: /usr/local/scaligent/install/install_config
    pgversion: 11
    layout_file: /tmp/hadoop_config.json
    no_sudo: 1
    minimal_sudo_install: 1
    offline: 1
    skip_r: 1
    skip_local_user_creation: 1

Related information

Was this page helpful?
×