For best results when implementing a high-availability solution using multiple instances of
VSP,
implement a round-robin solution with online-offline status using a load balancer.
- The load should be distributed evenly between available VSP instances.
- The VSP instances should all point to the
VMP Server, or to an internal
VMP load balancer if
VMP has been implemented in a cluster deployment.
If you are implementing the Vocera Smartphone Proxy in a high-availability solution,
these tests will be of use to you.
- http://<serverid>/vsptest: This query checks the status of the
VSP.
- http://<serverid>/WIC?query=test: This query is terminated on the
VMP Server, and tests the
status of VMP
and the connection between the VSP
and the VMP Server.
<serverid> is the
VSP IP address or domain name.
The interval between tests using http://<serverid>/WIC?query=test should be identical to the timeout interval on the
VSP (this is normally 30 seconds).
If two consecutive queries fail on a particular instance of
VSP:
- This VSP instance should be considered unavailable.
- The load balancer should route traffic to the other available VSP
instances.
-
An email notification should be sent to the administrator, along with the results of
the http://<serverid>/vsptest query. If this query fails, this VSP
service is down and should be considered offline, and the administrator needs to attend to the server (for example,
restart the server or start up the virtual machine). If this query passes, it may indicate that either the
VMP Server is unavailable or the connection to it is
unavailable. In this case, the administrator should look at the VMP Server
or the connection between it and the VSP.
http://<serverid>/WIC?query=test queries should continue. If a query passes, this instance of
VSP is again functional and can be considered online. The load
balancer should redistribute loading to this VSP instance.
Note: Optional smartphone URL filtering on the external load balancer can reduce loading on the
VSP instances.