How to Plan for Switch Clustering
Anticipating conflicts and compatibility issues is a high priority when you manage several switches through a cluster. This section describes the guidelines, requirements, and caveats that you should understand before you create the cluster:
Refer to the release notes for the list of Catalyst switches eligible for switch clustering, including which ones can be cluster command switches and which ones can only be cluster member switches, and for the required software versions and browser and Java plug-in configurations.
Automatic Discovery of Cluster Candidates and Members
The cluster command switch uses Cisco Discovery Protocol (CDP) to discover cluster member switches, candidate switches, neighboring switch clusters, and edge devices across multiple VLANs and in star or cascaded topologies.
Note Do not disable CDP on the cluster command switch, on cluster members, or on any cluster-capable switches that you might want a cluster command switch to discover. For more information about CDP, see Chapter33, “Configuring CDP”
Following these connectivity guidelines ensures automatic discovery of the switch cluster, cluster candidates, connected switch clusters, and neighboring edge devices:
Discovery Through CDP Hops
By using CDP, a cluster command switch can discover switches up to seven CDP hops away (the default is three hops) from the edge of the cluster. The edge of the cluster is where the last cluster member switches are connected to the cluster and to candidate switches. For example, cluster member switches 9 and 10 in Figure 6-1 are at the edge of the cluster.
In Figure 6-1, the cluster command switch has ports assigned to VLANs 16 and 62. The CDP hop count is three. The cluster command switch discovers switches 11, 12, 13, and 14 because they are within three hops from the edge of the cluster. It does not discover switch 15 because it is four hops from the edge of the cluster.
Figure 6-1 Discovery Through CDP Hops
Discovery Through Non-CDP-Capable and Noncluster-Capable Devices
If a cluster command switch is connected to a non-CDP-capable third-party hub (such as a non-Cisco hub), it can discover cluster-enabled devices connected to that third-party hub. However, if the cluster command switch is connected to a noncluster-capable Cisco device, it cannot discover a cluster-enabled device connected beyond the noncluster-capable Cisco device.
Figure 6-2 shows that the cluster command switch discovers the switch that is connected to a third-party hub. However, the cluster command switch does not discover the switch that is connected to a Catalyst 5000 switch.
Figure 6-2 Discovery Through Non-CDP-Capable and Noncluster-Capable Devices
Discovery Through Different VLANs
If the cluster command switch is a Catalyst 2970, Catalyst 3550, Catalyst 3560, or Catalyst 3750 switch, the cluster can have cluster member switches in different VLANs. As cluster member switches, they must be connected through at least one VLAN in common with the cluster command switch. The cluster command switch in Figure 6-3 has ports assigned to VLANs 9, 16, and 62 and therefore discovers the switches in those VLANs. It does not discover the switch in VLAN 50. It also does not discover the switch in VLAN 16 in the first column because the cluster command switch has no VLAN connectivity to it.
Catalyst 2900 XL, Catalyst 2950, and Catalyst 3500 XL cluster member switches must be connected to the cluster command switch through their management VLAN. For information about discovery through management VLANs, see the “Discovery Through Different Management VLANs” section. For more information about VLANs, see Chapter18, “Configuring VLANs”
Figure 6-3 Discovery Through Different VLANs
Discovery Through Different Management VLANs
Catalyst 2970, Catalyst 3550, Catalyst 3560, or Catalyst 3750 cluster command switches can discover and manage cluster member switches in different VLANs and different management VLANs. As cluster member switches, they must be connected through at least one VLAN in common with the cluster command switch. They do not need to be connected to the cluster command switch through their management VLAN. The default management VLAN is VLAN 1.
Note If the switch cluster has a Catalyst 3750 or 2975 switch or has a switch stack, that switch or switch stack must be the cluster command switch.
The cluster command switch and standby command switch in Figure 6-5 (assuming they are Catalyst 2960, Catalyst 2970, Catalyst 2975, Catalyst 3550, Catalyst 3560, or Catalyst 3750 cluster command switches) have ports assigned to VLANs 9, 16, and 62. The management VLAN on the cluster command switch is VLAN 9. Each cluster command switch discovers the switches in the different management VLANs except these:
- Switches 7 and 10 (switches in management VLAN 4) because they are not connected through a common VLAN (meaning VLANs 62 and 9) with the cluster command switch
- Switch 9 because automatic discovery does not extend beyond a noncandidate device, which is switch 7
Discovery Through Routed Ports
Note The LAN Base image supports static routing and RIP.
If the cluster command switch has a routed port (RP) configured, it discovers only candidate and cluster member switches in the same VLAN as the routed port.
The Layer 3 cluster command switch in Figure 6-4 can discover the switches in VLANs 9 and 62 but not the switch in VLAN 4. If the routed port path between the cluster command switch and cluster member switch 7 is lost, connectivity with cluster member switch 7 is maintained because of the redundant path through VLAN 9.
Figure 6-4 Discovery Through Routed Ports
Figure 6-5 Discovery Through Different Management VLANs with a Layer 3 Cluster Command Switch
Discovery of Newly Installed Switches
To join a cluster, the new, out-of-the-box switch must be connected to the cluster through one of its access ports. An access port (AP) carries the traffic of and belongs to only one VLAN. By default, the new switch and its access ports are assigned to VLAN 1.
When the new switch joins a cluster, its default VLAN changes to the VLAN of the immediately upstream neighbor. The new switch also configures its access port to belong to the VLAN of the immediately upstream neighbor.
The cluster command switch in Figure 6-6 belongs to VLANs 9 and 16. When new cluster-capable switches join the cluster:
- One cluster-capable switch and its access port are assigned to VLAN 9.
- The other cluster-capable switch and its access port are assigned to management VLAN 16.
Figure 6-6 Discovery of Newly Installed Switches
IP Addresses
You must assign IP information to a cluster command switch. You can assign more than one IP address to the cluster command switch, and you can access the cluster through any of the command-switch IP addresses. If you configure a cluster standby group, you must use the standby-group virtual IP address to manage the cluster from the active cluster command switch. Using the virtual IP address ensures that you retain connectivity to the cluster if the active cluster command switch fails and that a standby cluster command switch becomes the active cluster command switch.
If the active cluster command switch fails and the standby cluster command switch takes over, you must either use the standby-group virtual IP address or any of the IP addresses available on the new active cluster command switch to access the cluster.
You can assign an IP address to a cluster-capable switch, but it is not necessary. A cluster member switch is managed and communicates with other cluster member switches through the command-switch IP address. If the cluster member switch leaves the cluster and it does not have its own IP address, you must assign an IP address to manage it as a standalone switch.
For more information about IP addresses, see Chapter4, “Performing Switch Setup Configuration”
Hostnames
You do not need to assign a hostname to either a cluster command switch or an eligible cluster member. However, a hostname assigned to the cluster command switch can help to identify the switch cluster. The default hostname for the switch is Switch.
If a switch joins a cluster and it does not have a hostname, the cluster command switch appends a unique member number to its own hostname and assigns it sequentially as each switch joins the cluster. The number means the order in which the switch was added to the cluster. For example, a cluster command switch named eng-cluster could name the fifth cluster member eng-cluster-5.
If a switch has a hostname, it retains that name when it joins a cluster and when it leaves the cluster.
If a switch received its hostname from the cluster command switch, was removed from a cluster, was then added to a new cluster, and kept the same member number (such as 5), the switch overwrites the old hostname (such as eng-cluster-5) with the hostname of the cluster command switch in the new cluster (such as mkg-cluster-5). If the switch member number changes in the new cluster (such as 3), the switch retains the previous name ( eng-cluster-5).
Passwords
You do not need to assign passwords to an individual switch if it will be a cluster member. When a switch joins a cluster, it inherits the command-switch password and retains it when it leaves the cluster. If no command-switch password is configured, the cluster member switch inherits a null password. Cluster member switches only inherit the command-switch password.
If you change the member-switch password to be different from the command-switch password and save the change, the switch is not manageable by the cluster command switch until you change the member-switch password to match the command-switch password. Rebooting the member switch does not revert the password back to the command-switch password. We recommend that you do not change the member-switch password after it joins a cluster.
For more information about passwords, see the “Prevention for Unauthorized Switch Access” section.
For password considerations specific to the Catalyst 1900 and Catalyst 2820 switches, refer to the installation and configuration guides for those switches.
SNMP Community Strings
A cluster member switch inherits the command-switch first read-only (RO) and read-write (RW) community strings with @esN appended to the community strings:
- command-switch-readonly-community-string @ esN, where N is the member-switch number.
- command-switch-readwrite-community-string @ esN, where N is the member-switch number.
If the cluster command switch has multiple read-only or read-write community strings, only the first read-only and read-write strings are propagated to the cluster member switch.
The switches support an unlimited number of community strings and string lengths. For more information about SNMP and community strings, see Chapter37, “Configuring SNMP”
For SNMP considerations specific to the Catalyst 1900 and Catalyst 2820 switches, refer to the installation and configuration guides specific to those switches.
TACACS+ and RADIUS
If TACACS+ is configured on a cluster member, it must be configured on all cluster members. Similarly, if RADIUS is configured on a cluster member, it must be configured on all cluster members.The same switch cluster cannot have some members configured with TACACS+ and other members configured with RADIUS.
For more information about TACACS+, see the “Configuring TACACS+” section. For more information about RADIUS, see the “Configuring Radius Server Communication” section.
LRE Profiles
A configuration conflict occurs if a switch cluster has Long-Reach Ethernet (LRE) switches that use both private and public profiles. If one LRE switch in a cluster is assigned a public profile, all LRE switches in that cluster must have that same public profile. Before you add an LRE switch to a cluster, make sure that you assign it the same public profile used by other LRE switches in the cluster.
A cluster can have a mix of LRE switches that use different private profiles.
Managing Switch Clusters
Using the CLI to Manage Switch Clusters
You can configure cluster member switches from the CLI by first logging into the cluster command switch. Enter the rcommand user EXEC command and the cluster member switch number to start a Telnet session (through a console or Telnet connection) and to access the cluster member switch CLI. The command mode changes, and the Cisco IOS commands operate as usual. Enter the exit privileged EXEC command on the cluster member switch to return to the command-switch CLI.
This example shows how to log into member-switch 3 from the command-switch CLI:
If you do not know the member-switch number, enter the show cluster members privileged EXEC command on the cluster command switch. For more information about the rcommand command and all other cluster commands, refer to the switch command reference.
The Telnet session accesses the member-switch CLI at the same privilege level as on the cluster command switch. The Cisco IOS commands then operate as usual. For instructions on configuring the switch for a Telnet session, see the “Disabling Password Recovery” section.
Catalyst 1900 and Catalyst 2820 CLI Considerations
If your switch cluster has Catalyst 1900 and Catalyst 2820 switches running standard edition software, the Telnet session accesses the management console (a menu-driven interface) if the cluster command switch is at privilege level 15. If the cluster command switch is at privilege level 1 to 14, you are prompted for the password to access the menu console.
Command-switch privilege levels map to the Catalyst 1900 and Catalyst 2820 cluster member switches running standard and Enterprise Edition Software as follows:
- If the command-switch privilege level is 1 to 14, the cluster member switch is accessed at privilege level 1.
- If the command-switch privilege level is 15, the cluster member switch is accessed at privilege level 15.
Note The Catalyst 1900 and Catalyst 2820 CLI is available only on switches running Enterprise Edition Software.
For more information about the Catalyst 1900 and Catalyst 2820 switches, refer to the installation and configuration guides for those switches.
Using SNMP to Manage Switch Clusters
When you first power on the switch, SNMP is enabled if you enter the IP information by using the setup program and accept its proposed configuration. If you did not use the setup program to enter the IP information and SNMP was not enabled, you can enable it as described in the Chapter37, “Configuring SNMP” On Catalyst 1900 and Catalyst 2820 switches, SNMP is enabled by default.
When you create a cluster, the cluster command switch manages the exchange of messages between cluster member switches and an SNMP application. The cluster software on the cluster command switch appends the cluster member switch number ( @esN, where N is the switch number) to the first configured read-write and read-only community strings on the cluster command switch and propagates them to the cluster member switch. The cluster command switch uses this community string to control the forwarding of gets, sets, and get-next messages between the SNMP management station and the cluster member switches.
Note When a cluster standby group is configured, the cluster command switch can change without your knowledge. Use the first read-write and read-only community strings to communicate with the cluster command switch if there is a cluster standby group configured for the cluster.
If the cluster member switch does not have an IP address, the cluster command switch redirects traps from the cluster member switch to the management station, as shown in Figure 6-7. If a cluster member switch has its own IP address and community strings, the cluster member switch can send traps directly to the management station, without going through the cluster command switch.
If a cluster member switch has its own IP address and community strings, they can be used in addition to the access provided by the cluster command switch. For more information about SNMP and community strings, see Chapter37, “Configuring SNMP”
Figure 6-7 SNMP Management for a Cluster