Instance Count for Shared Services after ArcGIS Monitor Upgrade 2024.1 to 2025.0

[KoltWright]

After upgrading from ArcGIS Monitor 2024.1 to 2025.0, several of my shared pool services are now showing 0 min and 0 max instances. This has caused Instance Saturation Alerts to be created for about 45 of my services between my dev and prod environments. Not all shared services are experiencing this issue, though.

I have tried changing the pooling of a few services, reconnecting agents, and a few other methods to update the inaccurate service definition. For context, both my servers have 8 and 4 shared instances each. 

[KoltWright]

Update: if anyone was curious, after speaking with ESRI, it sounds like this is a known bug that will be resolved, likely in a later release.

[DavidColey · Accepted solution]

Yes I reported on this back on June 9

https://community.esri.com/t5/arcgis-monitor-questions/monitor-2025-map-image-services-in-shared-ins...

and I am still working with ESRI on the issue, but my support person has not told me yet that this is a known bug - they are only telling me that they are continuing to work on it.

In the meantime, the post does contain a workaround, where you can go into ArcGIS Server Administrator Directory and change the max instance value on map image service that is in the shared instance to something other than 0, (like 1 or 2) and the saturation value will resolve.

But I don't like leaving my services like this because ArcGIS Server Manager still reads 0,0 for my min and max instance counts (as it always has for map server services) while in the Admin directory my counts are now 0,1.  

This does not cause any stability issues for the services or other bad behavior - so far.

@DerekLaw 

[KoltWright]

@DavidColey, thanks for your reply! I must have missed your original post. I approached the solution in a slightly different manner, but it was probably not the best. For all the services that I had, with zero minimum and zero maximum, I disabled instance saturation reporting. I figured I could rely on the server instance saturation for now. I might implement your provided solution instead, though.