Google Search Appliance Configuring GSA Mirroring version 7.0 User Manual
Page 7
Google Search Appliance: Configuring GSA Mirroring
7
•
To the master search appliance only, in an active-passive configuration. In this configuration, you
can manually fail over search traffic to a replica node if the master search appliance fails.
•
To all nodes, in an active-active configuration. In this configuration, all nodes can serve results. You
use a load balancer to distribute search requests among the nodes. Ensure that you use a load
balancer that supports sticky user sessions, so that the user receives consistent search results.
The process for setting up mirroring on the search appliance is same for both cases. Load balancers are
typically used to achieve either configuration.
When the Master Search Appliance Fails in an
Active-Passive Configuration
All nodes should be monitored externally and when the master node is not responding, all search traffic
should be redirected to the replica node. You can redirect traffic by using a load-balancer, switching the
value of the DNS alias to the replica, or using an external application. Users will be reprompted for
credentials on the replica node. Changes to user-facing features such as User Added Results will not
propagate to other nodes from a replica node. A replica node can be promoted to be a master node,
and the new master node can resume crawling and start accepting feeds. For more information on how
to do this, see “Promoting a Replica Node to Master” on page 14. For more information on monitoring,
see the section “Setting up monitoring” in the article Design a search solution (
).
When a Search Appliance Fails in an Active-Active
Configuration
All nodes should be monitored externally. When a node in the configuration fails, your response
depends on how the load balancer works.
•
If the load balancer redirects traffic automatically from the failed node to nodes that are still
running, you do not need to take action.
•
If the load balancer does not redirect traffic automatically, you must configure the load balancer so
that search traffic is correctly directed to the running nodes.
When users with existing sessions are directed to a new node, they might have to log in again.
Settings that are Copied or Not Copied to the Replica
Search Appliances
The following settings and data are copied from the master to the replica search appliances:
•
Spelling data
•
Connector Manager definitions and configurations
•
Configuration data except the settings listed below
The following settings and data are not copied from the master to the replica search appliances:
•
Kerberos
•
SAML