We are planning on migrating from Server 10.0 (single machine deployment) to Server 10.1 (2 machine deployment). Currently we have all of our service on a single machine...which exposes the REST endpoints for them as (for example):
Server1 is the name of the server we have ArcGIS Server 10.0 installed on...and have all of the .msd's located at. We are planning on updating to 10.1 and adding another server machine (Server2). We also plan on installing the Web Adaptor on a separate machine. I have 2 questions:
1. Assuming we are going to use the 2 servers to help balance the load, should we place all of the map services on both machines(Server1 and Server2)? I know you can do clusters and create one cluster for geoprocessing and one for map services (for example), but we only have map services right now. So would placing the exact same msd's on both machines be the methodology for balancing the load?
2. Currently we have a flex website that consumes our REST service endpoints, which have our current computer name in it (server1'>http://server1/ArcGIS/rest/services/Map/Parcels/MapServer). If we go to 2 server machines, does that mean we will have 2 different REST endpoints for the same service?
If so, will installing the web adapator take care of this for us (i.e. consolidate the 2 REST endpoints into a single REST endpoint)? Any help would be much appreciated!
At 10.1, an ArcGIS Server site is composed of one or more GIS Server machines. Each GIS server in the site has some intelligence about what the other GIS servers are doing. Therefore, the site is able to balance the load among the available machines. This also makes the site elastic, such that when you add or remove a machine, the site can automatically adapt.
You can access any service directly from any GIS server machine in the site. So if you have two machines in the site, there are potentially two URLs you could use to access a service. However, in practice, most people will use one URL. Production deployments of ArcGIS Server should use the Web Adaptor, which provides a single point of entry for your GIS server. This is what you would reference in your Flex apps, etc. The Web Adaptor passes the requests on to the GIS servers.
When you publish a service, a copy of your map is placed in the server directories (which are located at a path you specify when you create the site). See these deployment scenario diagrams for options of where you can put the server directories and your GIS datasets: http://resources.arcgis.com/en/help/main/10.1/0154/015400000488000000.htm
A few questions which I hope you can help clarifying.
1. Using Steve's configuration, is it correct to expect that when a map service, MS1, is published on Server1, the map service is known to Server2 't without been published on Server2?
2. If I have map service MS1 published on Server1 and map service MS2 published on Server2, when I display the REST service directory of either server, both MS1 and MS2 are displayed?
