OpenText Analytics Database 26.2.x – Before you install Vertica

Setup: Platform and hardware requirements and recommendations

Mon, 01 Jan 0001 00:00:00 +0000

Hardware recommendations

The Vertica Analytics Platform is based on a massively parallel processing (MPP), shared-nothing architecture, in which the query processing workload is divided among all nodes of the Vertica database. OpenText highly recommends using a homogeneous hardware configuration for your Vertica cluster; that is, each node of the cluster should be similar in CPU, clock speed, number of cores, memory, and operating system version.

Note that OpenText has not tested Vertica on clusters made up of nodes with disparate hardware specifications. While it is expected that a Vertica database would functionally work in a mixed hardware configuration, performance will be limited to that of the slowest node in the cluster.

Vertica performs best on processors with higher clock frequency. When possible, choose a faster processor with fewer cores as opposed to a slower processor with more cores.

Tests performed both internally and by customers have shown performance differences between processor architectures even when accounting for differences in core count and clock frequency. When possible, compare platforms by installing Vertica and running experiments using your data and workloads. Consider testing on cloud platforms that offer VMs running on different processor architectures, even if you intend to deploy your Vertica database on premises.

Detailed hardware recommendations are available in Recommendations for Sizing Vertica Nodes and Clusters.

Platform requirements and recommendations

You must verify that your servers meet the platform requirements described in Supported Platforms. The Supported Platforms topics detail supported versions for the following:

OS for Server and Management Console (MC)
Supported Browsers for MC
Supported File Systems

Important

Deploy Vertica as the only active process on each host—other than Linux processes or software explicitly approved by Vertica. Vertica cannot be co-located with other software. Remove or disable all non-essential applications from cluster hosts.

Install the latest vendor-specific system software

Install the latest vendor drivers for your hardware.

Data storage recommendations

All internal drives connect to a single RAID controller.
The RAID array should form one hardware RAID device as a contiguous /data volume.

Install Perl

Before you perform the cluster installation, install Perl 5 on all the target hosts. Perl is available for download from www.perl.org.

Validation utilities

Vertica provides several validation utilities that validate the performance on prospective hosts. The utilities are installed when you install the Vertica RPM, but you can use them before you run the install_vertica script. See Validation scripts for more details on running the utilities and verifying that your hosts meet the recommended requirements.

Verify sudo

Vertica uses the sudo command during installation and some administrative tasks. Ensure that sudo is available on all hosts with the following command:

# which sudo
/usr/bin/sudo

If sudo is not installed, on all hosts, follow the instructions in How to Enable sudo on Red Hat Enterprise Linux.

When you use sudo to install Vertica, the user that performs the installation must have privileges on all nodes in the cluster.

Configuring sudo with privileges for the individual commands can be a tedious and error-prone process; thus, the Vertica documentation does not include every possible sudo command that you can include in the sudoers file. Instead, Vertica recommends that you temporarily elevate the sudo user to have all privileges for the duration of the install.

Note

See the sudoers and visudo man pages for the details on how to write/modify a sudoers file.

To allow root sudo access on all commands as any user on any machine, use visudo as root to edit the /etc/sudoers file and add this line:

## Allow root to run any commands anywhere
root   ALL=(ALL) ALL

After the installation completes, remove (or reset) sudo privileges to the pre-installation settings.

BASH shell requirements

All shell scripts included in Vertica must run under the BASH shell. If you are on a Debian system, then the default shell can be DASH. DASH is not supported. Change the shell for root and for the dbadmin user to BASH with the chsh command.

For example:

# getent passwd | grep root
root:x:0:0:root:/root:/bin/dash

# chsh
Changing shell for root.
New shell [/bin/dash]: /bin/bash
Shell changed.

Then, as root, change the symbolic link for /bin/sh from /bin/dash to /bin/bash:

# rm /bin/sh
# ln -s /bin/bash /bin/sh

Log out and back in for the change to take effect.

Setup: Communal storage for on-premises Eon Mode databases

Mon, 01 Jan 0001 00:00:00 +0000

If you create an Eon Mode database, you must plan for your use of communal storage to store your database's data. Communal storage is based on a shared storage, such as AWS S3 or Pure Storage FlashBlade servers.

Whatever communal storage platform you use, you must ensure that it is durable (protected against data loss). The data in your Eon Mode database is only as safe as the communal storage that contains it. Most cloud provider's object stores come with a guaranteed redundancy to prevent data loss. When you install an Eon Mode database on-premises, you may have to take additional steps to prevent data loss.

Planning communal storage capacity for on-premises databases

Most cloud providers do not limit the amount of data you can store in their object stores. The only real limit is your budget; storing more data costs more money.

When you create an Eon Mode database on-premises, your storage is limited to the size of your communal storage. Unlike the cloud, you must plan ahead for the amount of storage you will need. For example, if you have a Pure Admin FlashBlade installation with three 8TB blades, then in theory, your database can grow up to 24TB. In practice, you need to account other uses of your object store, as well as factors such as data compression, and space consumed by unreaped ROS containers (storage containers no longer used by Vertica but not yet deleted by the object store).

The following calculator helps you determine the size for your communal storage needs, based on your estimated data size and additional uses of your communal storage. The values with white backgrounds in the Value column are editable. Change them to reflect your environment.

Note

The calculator currently does not work in mobile browsers. Please use a desktop browser to view the calculator.

Setup: Configure the network

Mon, 01 Jan 0001 00:00:00 +0000

This group of steps involve configuring the network. These steps differ depending on your installation scenario. A single node installation requires little network configuration, because the single instance of the Vertica server does not need to communication with other nodes in a cluster. For cluster install scenarios, you must make several decisions regarding your configuration.

Vertica supports server configuration with multiple network interfaces. For example, you might want to use one as a private network interface for internal communication among cluster hosts (the ones supplied via the --hosts option to install_vertica) and a separate one for client connections.

Important

Vertica performs best when all nodes are on the same subnet and have the same broadcast address for one or more interfaces. A cluster that has nodes on more than one subnet can experience lower performance due to the network latency associated with a multi-subnet system at high network utilization levels.

Important notes

Network configuration is exactly the same for single nodes as for multi-node clusters, with one special exception. If you install Vertica on a single host machine that is to remain a permanent single-node configuration (such as for development or Proof of Concept), you can install Vertica using localhost or the loopback IP (typically 127.0.0.1) as the value for --hosts. Do not use the hostname localhost in a node definition if you are likely to add nodes to the configuration later.
If you are using a host with multiple network interfaces, configure Vertica to use the address which is assigned to the NIC that is connected to the other cluster hosts.
Use a dedicated gigabit switch. If you do not performance could be severely affected.
Do not use DHCP dynamically-assigned IP addresses for the private network. Use only static addresses or permanently-leased DHCP addresses.

Choose IPv4 or IPv6 addresses for host identification and communications

Vertica supports using either IPv4 or IPv6 IP addresses for identifying the hosts in a database cluster. Vertica uses a single address to identify a host in the database cluster. All the IP addresses used to identify hosts in the cluster must use the same IP family.

The hosts in your database cluster can have both IPv4 and IPv6 network addresses assigned to them. Only one of these addresses is used to identify the node within the cluster. You can use the other addresses to handle client connections or connections to other systems.

You tell Vertica which address family to use when you install it. By default, Vertica uses IPv4 addresses for hosts. If you want the nodes in your database to use IPv6 addresses, add the --ipv6 option to the arguments you pass to the install_vertica script.

Note

You cannot change the address family a database cluster uses after you create it. For example, suppose you created a Vertica database using IPv4 addresses to identify the hosts in your cluster. Then you cannot later change the hosts to use an IPv6 address for internal communications.

In most cases, the address family you select does not impact how your database functions. However, there are a few exceptions:

Use IPv4 addresses to identify the nodes in your cluster if you want to use the Management Console to manage your database. Currently, the MC does not support databases that use IPv6 addresses.
If you select IPv6 addressing for your cluster, it automatically uses point-to-point networking mode.
Currently, AWS is the only cloud platform on which Vertica supports IPv6 addressing. To use IPv6 on AWS, you must identify cluster hosts using IP addresses instead of host names. The AWS DNS does not support resolving host names to IPv6.
If you only assign IPv6 addresses to the hosts in your database cluster, you may have problems interfacing to other systems that do not support IPv6.

Part of the information you pass to the install script is the list of hosts it will use to form the Vertica cluster. If you use host names in this list instead of IP addresses, ensure that the host names resolve to the IP address family you want to use for your cluster. For example, if you want your cluster to use IPv6 addresses, ensure your DNS or /etc/hosts file resolves the host names to IPv6 addresses.

You can configure DNS to return both IPv4 and IPv6 addresses for a host name. In this case, the installer uses the IPv4 address unless you supply the --ipv6 argument. If you use /etc/hosts for host name resolution (which is the best practice), host names cannot resolve to both IPv4 and IPv6 addresses.

Optionally run spread on a separate control network

If your query workloads are network intensive, you can use the --control-network parameter with the install_vertica script (see Install Vertica with the installation script) to allow spread communications to be configured on a subnet that is different from other Vertica data communications.

The --control-network parameter accepts either the default value or a broadcast network IP address (for example, 192.168.10.255 ).

Configure SSH

Verify that root can use Secure Shell (SSH) to log in (ssh) to all hosts that are included in the cluster. SSH (SSH client) is a program for logging into a remote machine and for running commands on a remote machine.
If you do not already have SSH installed on all hosts, log in as root on each host and install it before installing Vertica. You can download a free version of the SSH connectivity tools from OpenSSH.
Make sure that /dev/pts is mounted. Installing Vertica on a host that is missing the mount point /dev/pts could result in the following error when you create a database:

TIMEOUT ERROR: Could not login with SSH. Here is what SSH said:Last login: Sat Dec 15 18:05:35 2007 from v_vmart_node0001

Allow passwordless SSH access for the dbadmin user

The dbadmin user must be authorized for passwordless ssh. In typical installs, you won't need to change anything; however, if you set up your system to disallow passwordless login, you'll need to enable it for the dbadmin user. See Enable secure shell (SSH) logins.

Setup: Operating system configuration overview

Mon, 01 Jan 0001 00:00:00 +0000

This topic provides a high-level overview of the OS settings required for Vertica. Each item provides a link to additional details about the setting and detailed steps on making the configuration change. The installer tests for all of these settings and provides hints, warnings, and failures if the current configuration does not meet Vertica requirements.

Before you install the operating system

The below sections detail system settings that must be configured when you install the operating system. These settings cannot be easily changed after the operating system is installed.

Configuration	Description
Supported Platforms	Verify that your servers meet the platform requirements described in Supported Platforms. Unsupported operating systems are detected by the installer. The installer generates one of the following issue identifiers if it detects an unsupported operating system: [S0320] - Fedora OS is not supported. [S0321] - The version of Red Hat/CentOS is not supported. [S0322] - The version of Ubuntu/Debian is not supported. [S0323] - The operating system could not be determined. The unknown operating system is not supported because it does not match the list of supported operating systems. [S0324] - The version of Red Hat is not supported.
LVM	Vertica Analytic Database supports Linux Volume Manager (LVM) on all supported operating systems. For information on LVM requirements and restrictions, see the section, Vertica Support for LVM.
File system	Choose the storage format type based on deployment requirements. Vertica recommends the following storage format types where applicable: ext3 ext4 NFS for backup XFS Amazon S3 Standard, Azure Blob Storage, or Google Cloud Storage for communal storage and related backup tasks when running in Eon Mode Note For the Vertica I/O profile, the ext4 file system is considerably faster than ext3. The storage format type at your backup and temporary directory locations must support fcntl lockf (POSIX) file locking.
Swap Space	A 2GB swap partition is required, regardless of the amount of RAM installed on your system. Larger swap space is acceptable, but unnecessary. Partition the remaining disk space in a single partition under "/". If you do not have the required 2GB swap partition, the installer reports this issue with identifier S0180. You typically define the swap partition when you install Linux. See your platform’s documentation for details on configuring the swap partition. Note Do not place a swap file on a disk containing the Vertica data files. If a host has only two disks (boot and data), put the swap file on the boot disk.
Disk Block Size	The disk block size for the Vertica data and catalog directories should be 4096 bytes, the default on ext4 and XFS file systems. You set the disk block size when you format your file system. If you change the block size, you will need to reformat the disk.
Memory	Vertica requires that your hosts have a minimum of 1GB of RAM per logical processor. If your hosts do not meet this requirement, the installer reports this issue with the identifier S0190. For performance reasons, you typically require more RAM than the minimum. In addition to the individual host RAM requirement, the installer also reports a hint if the hosts in your cluster do not have identical amounts of RAM. Ensuring your host have the same amount of RAM helps prevent performance issues if one or more nodes has less RAM than the other nodes in your database. Note In an Eon Mode database, after you create the initial cluster, you can configure subclusters that have different hardware specifications (including RAM) than the initial primary subcluster the installer creates. For more information on sizing your hardware, see the Vertica Knowledge Base Hardware documents.

Automatically configured operating system settings

These general OS settings are automatically made by the installer if they do not meet Vertica requirements. You can prevent the installer from automatically making these configuration changes by using the --no-system-configuration parameter for the install_vertica script.

For more information on each configuration setting, see Automatically configured operating system settings.

Configuration	Description
Nice Limits	The database administration user must be able to nice processes back to the default level of 0.
min_free_kbytes	The `vm.min_free_kbytes` setting in `/etc/sysctl.conf` must be configured sufficiently high. The specific value depends on your hardware configuration.
User Open Files Limit	The open file limit for the dbadmin user should be at least 1 file open per MB of RAM, 65536, or the amount of RAM in MB; whichever is greater.
System Open File Limits	The maximum number of files open on the system must not be less than at least the amount of memory in MB, but not less than 65536.
Pam Limits	/etc/pam.d/su must contain the line: `session required pam_limits.so` This allows for the conveying of limits to commands run with the `su -` command.
Address Space Limits	The address space limits (`as` setting) defined in /etc/security/limits.conf must be unlimited for the database administrator.
File Size Limits	The file sizelimits (`fsize` setting) defined in /etc/security/limits.conf must be unlimited for the database administrator.
User Process Limits	The `nproc` setting defined in /etc/security/limits.conf must be 1024 or the amount of memory in MB, whichever is greater.
Maximum Memory Maps	The vm.max_map_count in /etc/sysctl.conf must be 65536 or the amount of memory in KB / 16, whichever is greater.

Manually configured operating system settings

For more information on each configuration setting, see Manually configured operating system settings.

Configuration	Description
Disk Readahead	This disk readahead must be at least 2048, with a high of 8192. Set this high limit only with the help of Vertica support. The specific value depends on your hardware configuration.
NTP Services	The NTP daemon must be enabled and running, with the exception of Red Hat and CentOS systems.
SELinux	SElinux must be disabled or run in permissive mode.
CPU Frequency Scaling	Vertica recommends that you disable CPU Frequency Scaling. Important Your systems may use significantly more energy when CPU frequency scaling is disabled.
Transparent Hugepages	For Red Hat and CentOS, Transparent Hugepages must be set to `always`. For all other operating systems, Transparent Hugepages must be disabled or set to `madvise`.
I/O Scheduler	The I/O Scheduler for disks used by Vertica must be set to deadline or noop.
Support Tools	Several optional packages can be installed to assist Vertica support when troubleshooting your system.

System user requirements

The following tasks pertain to the configuration of the system user required by Vertica.

For more information on each configuration setting, see System user configuration.

Configuration	Required Setting(s)
System User Requirements	The installer automatically creates a user with the correct settings. If you specify a user with `--dba-use`r, then the user must conform to the requirements for the Vertica system user.
LANG Environment Settings	The LANG environment variable must be set and valid for the database administration user.
TZ Environment Settings	The TZ environment variable must be set and valid for the database administration user.

Setup: Automatically configured operating system settings

Mon, 01 Jan 0001 00:00:00 +0000

These general Operating System settings are automatically made by the installer. You can prevent the installer from automatically making these configuration changes by using the --no-system-configuration parameter for the install_vertica script.

Setup: Manually configured operating system settings

Mon, 01 Jan 0001 00:00:00 +0000

The topics in this section detail general Operating System settings that must be set manually.

Persisting operating system settings

To prevent manually set Operating System settings from reverting on reboot, you should configure some of these settings in the /etc/rc.local script. This script contains commands and scripts that run each time the system is booted.

Important

On reboot, SUSE systems use the /etc/init.d/after.local file rather than /etc/rc.local.

Vertica uses settings in /etc/rc.local to set the following functionality:

Editing /etc/rc.local

As the root user, open /etc/rc.local:
```
# vi /etc/rc.local
```
Enter a script or command. For example, to configure transparent hugepages to meet Vertica requirements, enter the following:
```
echo never > /sys/kernel/mm/redhat_transparent_hugepage/enabled
```
Important
On some Ubuntu/Debian systems, the last line in /etc/rc.local must be exit 0. All additions to /etc/rc.local must precede this line.
Save your changes, and close /etc/rc.local.
If you use Red Hat 7.0 or CentOS 7.0 or higher, run the following command as root or sudo:
```
$ chmod +x /etc/rc.d/rc.local
```

On reboot, the command runs during startup. You can also run the command manually as the root user, if you want it to take effect immediately.

Disabling tuning system service

If you use Red Hat 7.0 or CentOS 7.0 or higher, make sure the tuning system service does not start on when Vertica reboots. Turning off tuning prevents monitoring of your OS and any tuning of your OS based on this monitoring. Tuning also enables THP silently, which can cause issues in other areas such as read ahead.

Run the following command as sudo or root:

$ chkconfig tuned off

Setup: System user configuration

Mon, 01 Jan 0001 00:00:00 +0000

The following tasks pertain to the configuration of the system user required by Vertica.