In the previous blog post, How to Quickly Install Vertica on Large Clusters we saw that installation time of vertica reduced by 50% for a 3 node cluster. Now, with the release of 23.4 installation time can be reduced further.
The first step is to copy rpm to the same path on all vertica nodes using bash script. Best is to use /tmp or /home/dbadmin as all nodes have that path. Here in we have attached a small script which can copy RPM’s to all vertica nodes.
community.opentext.com/.../copy_5F00_rpm.txt
Usage: One needs to pass 2 arguments. First argument is the RPM file name and second argument is the path where the rpm needs to be copied.
Chmod +x copy_rpm.sh
sh copy_rpm.sh <rpm file name> <path to copy>
To use this feature, add --no-rpm-copy and --rpm-path flags with the install_vertica command along with options listed in earlier blog. When the --rpm-path is specified, there is no need to provide -r option.
Example:
First, run the RPM package installer using the below command.
sudo rpm -Uvh vertica-23.4.0-0.x86_64.RHEL6.rpm
Now run install_vertica using below command
[root@node1 dbadmin]# sudo /opt/vertica/sbin/install_vertica -s node1,node2,node3 --parallel-no-prompts --ssh-password '$vertica$' --no-rpm-copy --rpm-path /home/dbadmin/vertica-23.4.0-0.x86_64.RHEL6.rpm
Vertica Analytic Database 23.4.0-0 Installation Tool
Installing rpm in parallel on 2 hosts....
installing on: ['node2', 'node3']
Begin package installation on ['node2', 'node3']
End package installation on ['node2', 'node3']
node2 Package installation validated
node3 Package installation validated
install.log snippets:
2023-10-20 17:59:13.127 installer/150979:0x7f7ae53cc740 [install.install_RPMs] <INFO> Installing rpm in parallel on 2 hosts....
2023-10-20 17:59:13.139 installer/150979:0x7f7ae53cc740 [SSH.installNode] <INFO> Skipping rpm copy. Every node in the cluster is assumed to have its own rpm at /home/dbadmin/vertica-23.4.0-0. x86_64.RHEL6.rpm
The italicized log snippets above indicate that vertica is skipping the copy of the RPM while still doing parallel rpm installation. The installation took 40 seconds less time as compared to parallel installation. The higher the number of nodes, lesser the time it takes to complete installation with the skipping the rpm copy along with parallel installation.
For more information, see
