Cisco Redundancy Models for Cisco TMS
Cisco TMS redundancy configuration and overview
Page 20 of 32
Multiple TMS Servers, Load Balancing, and Clustered Database
The highest availability configuration is one that will provide automatic fail-over for both the Cisco TMS
Server and the SQL database. Automatic Fail-over for the SQL database is discussed in Section
0
Single Server, Clustered Database.
Combining a SQL Cluster with multiple Cisco TMS Servers being
fronted by a Network Load Balancer (NLB) will provide fully automatic fail-over for both Cisco TMS
and the SQL Server.
Managed devices must be configured with the IP address or DNS hostname of the Cisco TMS server
so those addresses must be ‘shared’ between all the Cisco TMS Servers. Simply pointing a unit to
multiple management servers is not sufficient as the management servers must be performing in
unison to not duplicate tasks and to share information properly. The Network Load Balancer performs
this ‘sharing’ function by forwarding
connections sent to these ‘shared’
addresses to a particular Cisco TMS
server from a pool of Cisco TMS
servers. Depending on the particular
NLB being used, the logic behind which Cisco TMS server the connection gets forwarded to can be
customized. The NLB must be able to forward all HTTP connections and SNMP Traps sent to the
‘shared’ addresses. Due to view state in web functionality, sticky connections should be enabled in the
load balancer to ensure the same web session stays on the same web server.
When the Cisco TMS servers need to initiate connections to managed devices, each server can
initiate the connection directly from the server to the managed device. Therefore it is required that the
individual Cisco TMS servers be directly addressable by their own IP address from the managed
systems, not just through the NLB.
In this deployment scenario, fail-over for Cisco TMS servers is handled by the forwarding logic of the
NLB. The NLB should know not to forward connections to a Cisco TMS server that is not responding.
Fail-over for the Cisco TMS database is handled automatically by the SQL Cluster.
Recovery Methods
Cisco TMS Server Failure.
The NLB should not forward connections to a failed Cisco TMS server, so fail-over to another
Cisco TMS server is automatic. The Failed server should be repaired and brought back online
when possible.
SQL Server Failure.
The SQL Cluster will automatically handle the fail-over of assigning the SQL Server resources to
another node in the cluster. This happens transparently to Cisco TMS. Please refer to the
Microsoft SQL Server documentation for information on restoring a node in a SQL Cluster
Installation
Domain Requirements.
Since you will be using multiple Cisco TMS servers, all Cisco TMS servers must be a member of
a domain, and all Cisco TMS users should be a member of that domain or a domain trusted by
the Cisco TMS server’s domain. Using Windows local user accounts is not supported when using
multiple Cisco TMS servers.
Setup SQL Cluster.
The SQL Cluster must be configured before installing Cisco TMS. Configuring a Microsoft SQL
Cluster is outside the scope of this document. Please refer to Microsoft’s documentation for
additional details
http://www.microsoft.com/technet/prodtechnol/sql/2000/maintain/failclus.mspx
Prior to installing Cisco TMS, the SQL Cluster should be active and must have the Virtual SQL Server
installed and active that you wish to use for Cisco TMS. The Virtual SQL Server will have an instance
name and IP Address to reach the SQL server and ‘mixed mode authentication’ must be enabled.
Installing primary Cisco TMS application.
Once the Virtual SQL Server is configured, Cisco TMS can be installed on the first Cisco TMS
server. The Cisco TMS installation program allows you to select an existing remote SQL server as
part of the custom installation. Select or enter the instance name or IP address of the Virtual SQL
Server. The installation program will fully handle the configuration of the database on the existing
Error! Objects cannot be created from editing field codes.
Figure 2 Network Load Balancing
