You can configure the HTTP server to make it highly available
and you can set system properties to alleviate the symptoms that result
from failover of the HTTP server.
Before you begin
When the HTTP server fails, the user can experience one of
the symptoms of failover. To alleviate these failover symptoms for
the HTTP server, you can opt to show a warning message when the server
connection is lost. In the System Properties application, set the
mxe.webclient.lostconnectionwarningonly property
to
true to enable the system message. To set the
intervals for showing system failure and reconnection messages, modify
the
mxe.webclient.lostconnectionrecheckinterval and
mxe.webclient.lostconnectionwarninginterval properties.
About this task
Users access the product application through the HTTP Service
IP or load balancer. This solution masks an HTTP server failover from
the user, and causes only a moment of inaccessibility while the Service
IP switches to the standby node.
Procedure
- Install IBM® HTTP Server on
two or more nodes. The first node is the active node and the subsequent
nodes are available to support failover. HTTP server installation
files can be installed locally or on a highly available shared disk
and mounted on the active node. You can use cluster management software
to configure and control mounting on a shared disk.
- Install cluster manager software , such as System
Automation for Multiplatforms,
on all nodes.
- To automate the monitoring of the HTTP service status and
startup and failover to standby nodes, create resources and policies
on the cluster manager. For information about creating
resources and policies, see the documentation for your cluster management
software product.
- If you do not have an external load balancer, apply a Service
IP to the active node by creating a resource and policy on your cluster
manager. Users can access the same product URL independent of which
system is active. If there is an external load balancer
to serve the client requests, then you can have multiple HTTP nodes
active at one time. No Service IP is required, because the external
load balancer distributes the requests across the HTTP servers.