Installation prerequisites

• python-2.6 with argparse
• python-2.7

Red Hat Enterprise Linux 7 must be loaded on all nodes that will be designated as protocol nodes. The installation toolkit and instructions within this document will not work with earlier versions of Red Hat Enterprise Linux nor with other vendor operating systems at this time.

• If you want to run protocols, CCR must be available.
• mmchconfig release=LATEST must be run.
• A GPFS cluster and file system are required. If this infrastructure does not already exist, you must install the GPFS software, create a GPFS cluster and create a file system. You can use the spectrumscale installation toolkit to do this. See Using the spectrumscale installation toolkit to perform installation tasks: Explanations and examples.
• The GPFS file system must be mounted on all GPFS protocol nodes.
• It is strongly recommended to configure the file system to only support NFSv4 ACLs. You can use the spectrumscale installation toolkit to do this if you use it to install GPFS. See Using the spectrumscale installation toolkit to perform installation tasks: Explanations and examples.

  Alternatively, you can use the -k nfs4 parameter for mmcrfs. NFSv4 ACLs are a requirement for ACL usage with the SMB and NFS protocols. For more information, see mmcrfs examples in the IBM Spectrum Scale: Administration and Programming Reference.

• Quotas are not enabled by default in GPFS file systems but are recommended for use with SMB and NFS protocols. For more information about quota management, see Enabling and disabling GPFS quota management in IBM Spectrum Scale: Administration and Programming Reference.

• For Object, the installation toolkit creates an independent fileset in the GPFS file system that you name.

You can use the spectrumscale installation toolkit to set up this shared root storage area if you use the toolkit for GPFS installation and file system creation. See Using the spectrumscale installation toolkit to perform installation tasks: Explanations and examples.

The node used for initiating installation of SMB, NFS, and/or Object protocols via the installation toolkit must be able to communicate via an internal or external network with all protocol nodes to be installed. All nodes also require SSH keys to be set up so that the installation toolkit can run remote commands without any prompts. Examples of common prompts are a prompt asking for a remote node's password and/or a prompt asking a Y/N question. No prompts should exist when using SSH among any cluster nodes to and from each other, and to and from the node designated for installation.

While reading through the following firewall port prerequisites, refer to this example for opening the Red Hat 7 firewall :

$ firewall-cmd --permanent --add-port=lowerlimit-upperlimit/tcp
$ firewall-cmd --permanent --add-port=lowerlimit-upperlimit/tcp
$ firewall-cmd --permanent --add-port=lowerlimit-upperlimit/udp
$ firewall-cmd --add-port=lowerlimit-upperlimit/tcp
$ firewall-cmd –add-port=lowerlimit-upperlimit/udp

After any firewall changes are done, on any node that these changes were done, the firewall service needs to be restarted (so you might want to do this once after applying all the relevant firewall changes).

$ systemctl restart firewalld.service

The installation toolkit uses TCP port 8889 for communicating with the Chef server. This port must be reachable from all protocol nodes. The installation toolkit does not change any SELinux settings; the user must change those settings, if required.

The SMB protocol uses external TCP port 445 and must be reachable by clients on all protocol nodes. The internal TCP port 4379 is used for internal communication between the protocol nodes and therefore must be reachable from all protocol nodes. If SMB is enabled and these ports are not open, the installation toolkit will fail during precheck. The failure will indicate to make these ports accessible and retry.

The Object protocol uses external TCP ports 5000 for the keystone server and 8080 for the proxy server. If Object is enabled, both must be reachable from all protocol nodes. Internal TCP ports used by object include 6200 (object server), 6201 (container server), 6202 (account server), 35357 (keystone), 11211 (memcached), 5431 (Object Postgres instance). All internal ports needed by the Object protocol must be reachable within the cluster from all protocol nodes.

GPFS cluster operation requires port 1191 be opened and reachable among all nodes within the internal GPFS cluster network. In addition, an ephemeral port range must be opened as well.

$ mmchconfig tscCmdPortRange=lowerlimit-upperlimit

Follow up by opening these ports on each node as well.

See the topics about GPFS port usage and firewall recommendations in the IBM Spectrum Scale: Advanced Administration Guide in IBM Spectrum Scale in IBM® Knowledge Center.

rhel-x86_64-server-7 Red Hat Enterprise Linux Server

For more information about Yum repositories, see Configuring Yum and Yum Repositories (https://access.redhat.com/documentation/en-US/Red_Hat_Enterprise_Linux/7/html/System_Administrators_Guide/sec-Configuring_Yum_and_Yum_Repositories.html).

It is recommended that Network Time Protocol (NTP) be configured on all nodes in your system to ensure that the clocks of all of the nodes are synchronized. Clocks that are not synchronized will cause debugging issues and authentication problems with the protocols.

However, if you are manually installing IBM Spectrum Scale or would prefer to manually configure NTP, see the following example of enabling NTP on a node.

# yum install -y ntp
# ntpdate <NTP_server_IP>
# systemctl enable ntpd
# systemctl start ntpd
# timedatectl list-timezones
# timedatectl set-timezone
# systemctl enable ntpd

You need to ensure that the boost-regex package is installed in the system before installation of the Performance Monitoring tool.

See the topic about changing configurations to enable spectrumscale to collect core dump data in Understanding the spectrumscale installation toolkit options.