Problems due to missing prerequisites

Use this information to ensure that prerequisites are met before using the installation toolkit for installation, deployment, and upgrade.

Passwordless SSH setup

You can manually verify and set up passwordless SSH as follows.
  1. Verify that passwordless SSH is set up using the following commands.
    ssh <host name of the first node>
ssh <host name of the second node>

    Repeat this on all nodes.

    Verify that the user can log into the node successfully without being prompted for any input and that there are no warnings.

    ssh <FQDN of the first node>
ssh <FQDN of the second node>

    Repeat this on all nodes.

    Verify that the user can log into the node successfully without being prompted for any input and that there are no warnings.

    ssh <IP address of the first node>
ssh <IP address of the second node>

    Repeat this on all nodes.

    Verify that the user can log into the node successfully without being prompted for any input and that there are no warnings.

  2. If needed, set up passwordless SSH using the following commands.
    Note: This is one of the several possible ways of setting up passwordless SSH.
    ssh-keygen

    Repeat this on all cluster nodes.

    ssh-copy-id <host name of the first node>
ssh-copy-id <host name of the second node>

    Repeat this on all nodes.

    ssh-copy-id <FQDN of the first node>
ssh-copy-id <FQDN of the second node>

    Repeat this on all nodes.

Yum repository setup

  • Verify that the Yum repository is set up using the following command on all cluster nodes.
    yum repolist

    This command should run clean with no errors if the Yum repository is set up.

Firewall configuration

It is recommended that firewalls are in place to secure all nodes. For more information, see Securing the IBM Spectrum Scale system using firewall.

  • If you need to open specific ports, use the following steps on Red Hat Enterprise Linux nodes.
    1. Check the firewall status.
      systemctl status firewalld
    2. Open ports required by the installation toolkit.
      firewall-cmd --permanent --add-port 8889/tcp
firewall-cmd --add-port 8889/tcp
firewall-cmd --permanent --add-port 10080/tcp
firewall-cmd --add-port 10080/tcp

CES IP address allocation

As part of the deployment process, the IBM Spectrum Scale checks routing on the cluster and applies CES IPs as aliases on each protocol node. Furthermore, as service actions or failovers, nodes dynamically lose the alias IPs as they go down and other nodes gain additional aliases to hold all of the IPs passed to them from the down nodes.

Example - Before deployment

The only address here is 192.168.251.161, which is the ssh address for the node. It is held by the eth0 adapter.

# ifconfig -a
eth0: flags=4163<UP,BROADCAST,RUNNING,MULTICAST>mtu 1500
        inet 192.168.251.161  netmask 255.255.254.0  broadcast 192.168.251.255
        inet6 2002:90b:e006:84:250:56ff:fea5:1d86  prefixlen 64  scopeid 0x0<global>
        inet6 fe80::250:56ff:fea5:1d86  prefixlen 64  scopeid 0x20<link>
        ether 00:50:56:a5:1d:86  txqueuelen 1000  (Ethernet)
        RX packets 1978638  bytes 157199595 (149.9 MiB)
        RX errors 0  dropped 2291  overruns 0  frame 0
        TX packets 30884  bytes 3918216 (3.7 MiB)
        TX errors 0  dropped 0 overruns 0  carrier 0  collisions 0	

# ip addr
2: eth0:<BROADCAST,MULTICAST,UP,LOWER_UP>mtu 1500 qdisc mq state UP qlen 1000
    link/ether 00:50:56:a5:1d:86 brd ff:ff:ff:ff:ff:ff
    inet 192.168.251.161/23 brd 192.168.251.255 scope global eth0
       valid_lft forever preferred_lft forever
    inet6 2002:90b:e006:84:250:56ff:fea5:1d86/64 scope global dynamic
       valid_lft 2591875sec preferred_lft 604675sec
    inet6 fe80::250:56ff:fea5:1d86/64 scope link
       valid_lft forever preferred_lft forever

Example - After deployment

Now that the CES IP addresses exist, you can see that aliases called eth0:0 and eth0:1 have been created and the CES IP addresses specific to this node have been tagged to it. This allows the ssh IP of the node to exist at the same time as the CES IP address on the same adapter, if necessary. In this example, 192.168.251.161 is the initial ssh IP. The CES IP 192.168.251.165 is aliased onto eth0:0 and the CES IP 192.168.251.166 is aliased onto eth0:1. This occurs on all protocol nodes that are assigned a CES IP address. NSD server nodes or any client nodes that do not have protocols installed on them do not get a CES IP.

Furthermore, as service actions or failovers, nodes dynamically lose the alias IPs as they go down and other nodes gain additional aliases such as eth0:1 and eth0:2 to hold all of the IPs passed to them from the down nodes.

# ifconfig -a
eth0: flags=4163<UP,BROADCAST,RUNNING,MULTICAST> mtu 1500
        inet 192.168.251.161  netmask 255.255.254.0  broadcast 192.168.251.255
        inet6 2002:90b:e006:84:250:56ff:fea5:1d86  prefixlen 64  scopeid 0x0<global>
        inet6 fe80::250:56ff:fea5:1d86  prefixlen 64  scopeid 0x20<link>
			  ether 00:50:56:a5:1d:86  txqueuelen 1000  (Ethernet)
        RX packets 2909840  bytes 1022774886 (975.3 MiB)
        RX errors 0  dropped 2349  overruns 0  frame 0
        TX packets 712595  bytes 12619844288 (11.7 GiB)
        TX errors 0  dropped 0 overruns 0  carrier 0  collisions 0
eth0:0: flags=4163<UP,BROADCAST,RUNNING,MULTICAST>mtu 1500
        inet 192.168.251.165  netmask 255.255.254.0  broadcast 192.168.251.255
        ether 00:50:56:a5:1d:86  txqueuelen 1000  (Ethernet)
eth0:1: flags=4163<UP,BROADCAST,RUNNING,MULTICAST>mtu 1500
        inet 192.168.251.166  netmask 255.255.254.0  broadcast 192.168.251.255
        ether 00:50:56:a5:1d:86  txqueuelen 1000  (Ethernet)

# ip addr
2: eth0: <BROADCAST,MULTICAST,UP,LOWER_UP>mtu 1500 qdisc mq state UP qlen 1000
    link/ether 00:50:56:a5:1d:86 brd ff:ff:ff:ff:ff:ff
    inet 192.168.251.161/23 brd 9.11.85.255 scope global eth0
       valid_lft forever preferred_lft forever
    inet 192.168.251.165/23 brd 9.11.85.255 scope global secondary eth0:0
       valid_lft forever preferred_lft forever
    inet 192.168.251.166/23 brd 9.11.85.255 scope global secondary eth0:1
       valid_lft forever preferred_lft forever
    inet6 2002:90b:e006:84:250:56ff:fea5:1d86/64 scope global dynamic
       valid_lft 2591838sec preferred_lft 604638sec
    inet6 fe80::250:56ff:fea5:1d86/64 scope link
       valid_lft forever preferred_lft forever

Addition of CES IPs to /etc/hosts

Although it is highly recommended that all CES IPs are maintained in a central DNS and that they are accessible using both forward and reverse DNS lookup, there are times when this might not be possible. IBM Spectrum Scale always verify that forward or reverse DNS lookup is possible. To satisfy this check without a central DNS server containing the CES IPs, you must add the CES IPs to /etc/hosts and create a host name for them within /etc/hosts. The following example shows how a cluster might have multiple networks, nodes, and IPs defined.

For example:

# cat /etc/hosts
127.0.0.1   localhost localhost.localdomain localhost4 localhost4.localdomain4
::1         localhost localhost.localdomain localhost6 localhost6.localdomain6
 
# These are external addresses for GPFS
# Use these for ssh in. You can also use these to form your GPFS cluster if you choose
198.51.100.2   ss-deploy-cluster3-1.example.com   ss-deploy-cluster3-1   
198.51.100.4   ss-deploy-cluster3-2.example.com   ss-deploy-cluster3-2
198.51.100.6   ss-deploy-cluster3-3.example.com   ss-deploy-cluster3-3
198.51.100.9   ss-deploy-cluster3-4.example.com   ss-deploy-cluster3-4

# These are addresses for the base adapter used to alias CES-IPs to.
# Do not use these as CES-IPs.
# You could use these for a gpfs cluster if you choose
# Or you could leave these unused as placeholders
203.0.113.7    ss-deploy-cluster3-1_ces.example.com   ss-deploy-cluster3-1_ces
203.0.113.10   ss-deploy-cluster3-2_ces.example.com   ss-deploy-cluster3-2_ces
203.0.113.12   ss-deploy-cluster3-3_ces.example.com   ss-deploy-cluster3-3_ces   
203.0.113.14   ss-deploy-cluster3-4_ces.example.com   ss-deploy-cluster3-4_ces   

# These are addresses to use for CES-IPs
203.0.113.17  ss-deploy-cluster3-ces.example.com   ss-deploy-cluster3-ces
203.0.113.20  ss-deploy-cluster3-ces.example.com   ss-deploy-cluster3-ces    
203.0.113.21  ss-deploy-cluster3-ces.example.com   ss-deploy-cluster3-ces
203.0.113.23  ss-deploy-cluster3-ces.example.com   ss-deploy-cluster3-ces

In this example, the first two sets of addresses have unique host names and the third set of addresses that are associated with CES IPs are not unique. Alternatively, you could give each CES IP a unique host name but this is an arbitrary decision because only the node itself can see its own /etc/hosts file. Therefore, these host names are not visible to external clients/nodes unless they too contain a mirror copy of the /etc/hosts file. The reason for containing the CES IPs within the /etc/hosts file is solely to satisfy the IBM Spectrum Scale CES network verification checks. Without this, in cases with no DNS server, CES IPs cannot be added to a cluster.

Parent topic: Resolving most frequent problems related to installation, deployment, and upgrade