Installation requirements for IBM Spectrum LSF Suite for Enterprise
Learn about special installation requirements for your IBM Spectrum LSF Suite for Enterprise cluster. The different roles that hosts in your cluster can take are also described.
Installation prerequisites
IBM Spectrum LSF Suite for Enterprise uses an RPM installation program.
The RPM installer for IBM Spectrum LSF Suite for Enterprise deploys a fully functional cluster with minimal user input. It also allows administrators to easily apply updates to the cluster.
- A computing environment that consists of two or more hosts
- A host that has at least one network interface to act as the LSF management host
- A supported operating system that is preinstalled on the LSF management host
- Compute hosts that can be set to boot over a network
The IBM Spectrum LSF Suite for Enterprise must meet the minimum hardware and software requirements.
Host prerequisites
LSF_Servers and
LSF_Clients role machines can be on a shared directory. For better performance, Elasticsearch components are always installed locally.
- LSF_Master role requires 2 GB in /opt/ibm
- LSF_Server role requires 1.5 GB in /opt/ibm
- LSF_Clients role requires 300 MB in /opt/ibm
- GUI_Hosts role requires 32 GB in /opt/ibm
- DB_Host role requires 150 MB in the root directory (/)
Large clusters should use SSDs for the database directory /var/lib/mysql and the Elasticsearch directory /opt/ibm/elastic/elasticsearch.
- /opt/ibm/lsfsuite/lsf/work contains job data.
- /opt/ibm/lsflogs contains LSF logs.
- /opt/ibm/elastic/elasticsearch contains Elasticsearch data and logs. This directory can get large.
- The high-availability shared directory (HA_shared_dir) contains the LSF work and configuration directories.
To achieve the highest degree of performance and scalability, use a powerful management host.
LSF has no minimum CPU requirement. For the systems that LSF is supported on, any host with sufficient physical memory can run LSF management host. Swap space is normally configured as twice the physical memory. LSF daemons in a cluster on Linux x86-64 use about 488 MB of memory when no jobs are running. Active jobs use most of the memory that LSF requires.
| Cluster size | Active jobs | Minimum required memory (typical) | Recommended server CPU |
|---|---|---|---|
| Small (<100 hosts) | 1,000 | 1 GB (32 GB) | Any server CPU |
| 10,000 | 2 GB (32 GB) | Recent server CPU | |
| Medium (100 - 1000 hosts) | 10,000 | 4 GB (64 GB) | Multi-core CPU (2 cores) |
| 50,000 | 8 GB (64 GB) | Multi-core CPU (4 cores) | |
| Large (>1000 hosts) | 50,000 | 16 GB (128 GB) | Multi-core CPU (4 cores) |
| 500,000 | 32 GB (256 GB) | Multi-core CPU (8 cores) |
| Cluster size | Active jobs | Minimum required memory (typical) | Recommended server CPU |
|---|---|---|---|
| Small (<100 hosts) | 1,000 | 16 GB (32 GB) | Multi-core CPU (4 cores) |
| Medium (100 - 1000 hosts) | 50,000 | 32 GB (64 GB) | Multi-core CPU (8 cores) |
| Large (>1000 hosts) | 500,000 | 96 GB (256 GB) | Multi-core CPU (16 cores) |
Operating system configuration
Installation on RHEL and CentOS has no operating system configuration prerequisites or internet connection. Installation on Ubuntu requires Ansible Version 2.2 and RPM tools and an internet connection to get any dependencies required during installation.
Correct host name resolution is essential. DNS or hosts files must be accurate and complete. All the hosts that are listed in the lsf-inventory file must be resolvable on all other hosts.
The operating system repository must be configured before installation so that installation dependency packages can be automatically pulled in. Each server must be able to use the OS repository. Each host must set passwordless SSH from the deployer host to enable the deployer host to log in to each host in the cluster with SSH so that Ansible can deploy the cluster.
The primary LSF administrator is lsfadmin. The primary administrator owns the LSF configuration files and log files for job events. If the lsfadmin user does not exist, the lsfadmin user account is created automatically as the default cluster administrator account. It is created as a system account that does not expire; it contains a default UID of 495 (for Fix Pack 12 and earlier) or 1000321495 (for Fix Pack 13 and later). The installation can also use an existing lsfadmin account that you created in LDAP or other system.
The high-availability (HA) installation requires a shared directory (either NFS or IBM Spectrum Scale). You can also configure an optional external database in the HA configuration.
- The following command returns the IP address of the
host:
host <host_name> - The following command returns the host
name:
host <host_IP> - Use the ping command with either the host name or host IP address to make sure that the host is reachable.
- Use the yum command to test the OS repository. The following command return
the proper OS yum
repository:
yum repolist - Run the yum update command. If you can't use yum
update, try any package name that is shipped in OS
distribution:
yum list mariadb - Check that the JDBC driver is
available:
yum list mysql-connector-java - Make sure that the global profile.d file is sourced by root on all hosts.
- Check that passwordless SSH is set up from deployer to all hosts:
ssh-copy-id <host_name>
Network configuration and host roles
A simple corporate network with one NIC that all servers are connected to is the most basic setup. Each node in the cluster must resolve both the management and deployment hosts.
You can also have a configuration with multiple networks. For example, the LSF servers might be connected to a low-latency network for running MPI workload over Infiniband, with a private network for handling LSF workload. LSF clients can be configured on the corporate network for the users to submit jobs from. The GUI hosts must be accessible on the corporate network. For installation, use the corporate network host name for the GUI hosts.
