Using a single machine in multiple clusters

[StephanKünster]

Hi all,

I'm wondering why it's impossible to use a registered machine in multiple clusters.

Here is where I come from:
I'm running an ArcGIS Server site (10.1.1) with two machines (GIS A, GIS B). After joining the site both GIS A and GIS B are running within 'default' cluster. Now I want to build two clusters: first cluster (1) should use GIS A and GIS B but second cluster (2) should only use GIS B. The reason is that I want to separate between "high available" (cluster 1) and "normal" (cluster 2) services. Services running within cluster 1 should run even if one of the machines becomes unavailable, but services in cluster 2 are less important and thus it wouldn't be a problem if they became unavailable. The advantage of this scenario is that I would save resources on GIS A because services running in cluster 2 wouldn't use CPU, RAM and disk space of GIS A as they are not published on GIS A.

In a second step one could imagine to restrict user access within the AGS security not only by groups but also by certain clusters, e.g. restrict one to only be able to use services hosted within cluster 2.

How would you implement this scenario right now and why is this functionality not implemented yet? Maybe there are good reasons and I'm missing something?

I'd find it very useful to have this functionality. What do you think?

Best regards,
Stephan

[divyamgulati1]

You are right!  it is impossible to have a registered machine in multiple clusters.
Still what you can do is install a VM on machine B, install arcgis server on it and join the site. May be this is how you can make a single machine available in two clusters..

[StephanKünster]

anyone else?

[JustinRodriguez]

Hello
This is impossible as only one install can connect to one Site (config-store/directories). For any suggestions or improvements, please ensure you submit your ideas to ideas.arcgis.com. Thank you very much-

Justin

[JustinRodriguez]

Hello,
To answer your question about high availability:

If you install servers in the same cluster, you achieve failover. Services in the same cluster will failover to any other machine in the cluster. Services will not failover to a machine in another cluster though.

If you install a machine in a new cluster, you achieve resource (RAM/CPU) load balancing. This is a manual process, as you must migrate or publish your services specifically in the cluster you want, but it will ensure that the load is distributed.

Here is some more information on clustering:

About GIS server clusters
http://resources.arcgis.com/en/help/main/10.1/index.html#//015400000418000000


Thanks-
Justin

[StevenGraf1]

If you install a machine in a new cluster, you achieve resource (RAM/CPU) load balancing. This is a manual process, as you must migrate or publish your services specifically in the cluster you want, but it will ensure that the load is distributed.

We recently upgraded to Server 10.2.2 and we have 2 machines in the same cluster for failover purposes.  At version 10.1.1 we did manually have to divide up our 2 servers to distribute the load.  It seems that in 10.2.2 both machines in the default cluster are sharing resources.  Is this correct?  I can't find much documentation for the release of 10.2.2.  They are using the exact same amount of memory and cpu power.  We currently don't have a network load balancer in place.