Hey there,
We're running ArcGIS GIS Server 11.2 and have recently made a change to our infrastructure that requires we use Replaced database connections on our production server, as we cannot access those databases from ArcGIS Pro on our PCs. We do have a dev/test server which uses a Referenced database connection.
When using Referenced data stores (i.e. dev/test) publishing services is quick (up to 3 minutes).
The same service to production, where the data is then Replaced, can take up to 20 minutes (5-6x as long).
Other functions, such as issuing a /changeProvider command via the REST API or rebuilding the cache also take a lot longer on the 'Replaced' environment.
We've spent some considerable time looking into this, we've had our database and system engineers check what the communication between GIS server and db cluster is, which isn't seeing any strain. Network and local factors we've excluded by running this from a VM located in the datacentre. Job logs don't show any errors, server logs show that it's communicating properly with the publishing machine.
So for me, it would be very interesting to know;
- Why is using a Replaced database connection this much slower than a Referenced connection?
We get no errors back from the functions, it just takes a really long time. There is little documentation on this, all we can find is: Data sources for ArcGIS Server—ArcGIS Server | Documentation for ArcGIS Enterprise
- Is this intentional behaviour of the 'Replaced' functionality or should we be reporting this as a bug?
Thanks!
