Clustering in Teiid
Since Teiid is installed in WildFly, there is no additional configuration needed beyond what was performed when Teiid is setup in Domain Mode. See the Domain Mode section in the Teiid Installation Guide. Just make sure that you installed Teiid in every WildFly node and started all WildFly instances in the Domain mode that to be a part of the cluster.
Typically users create clusters to improve the performance of the system through:
-
Load Balancing: Take look at HAProxy below and in the Client Developer’s Guide on how to use simple load balancing between multiple nodes.
-
Fail Over: Take look at the Client Developer’s Guide on how to use fail over between multiple nodes.
-
Distributed Caching: This is automatically done for you once you configure it as specified above.
-
Event distribution: metadata and data modifications will be distributed to all cluster members.
In the Domain mode, the only way a user can deploy any artifacts is using either CLI or using the Admin API. Copying VDB directly into the "deployments" directory is not supported.
|
Note
|
some load balancers have timeouts that cannot be adjusted. You may need to adjust the tcp_keepalive_time on your client OS. The default is typically 2 hours, which is much too long in many cases. See Custom Configuration of TCP Socket Keep-Alive Timeouts. |
HAProxy
HAProxy may be used for load-balancing and high availability. A good tutorial is located at Load Balancing JDV-HAProxy or see Luigi Fugaro’s example.
The load balancer should use an algorithm that supports sticky connections as Teiid sessions as specific to the original host. For HAProxy it is recommended that you use leastconn or source.