Calculate Cluster Key - Operation was attempted on empty geometry

AmberY

We encountered an error while running the Calculate Cluster Key tool: "Operation was attempted on empty geometry." In our case, the UNM implementation was utilising a custom coordinate system.

We then tested the tool on Water Distribution UN Foundation gdb, which uses Esri coordinate system (NAD 1983 StatePlane Illinois East FIPS 1201, US Feet), and the tool functioned as expected. However, after we changed the spatial reference of the Water Distribution UN Foundation gdb to the same custom coordinate system, it hit the same error.

Was wondering if this tool does not support the UNM implementation in custom coordinate system or if there is something i missed?

Our goal is to improve UNM performance in any way possible as it is noticeable that UNM tends to render slower in web app level. Any suggestion on how to improve UNM performance is very much appreciate as well!

Thanks in advance!

gis_KIWI4

@AmberY - I haven't come across this error before but my understanding is that the Cluster Keys help with topological operations like tracing. I do not believe they will make a big difference for the rendering of the UN in webmaps.

If you want to improve performance in the webmaps then the below may have the most improvement -

1) Using Simple symbols can improve rendering times
2) Indexing fields that are used to derive the symbology (eg if you are using Lifecycle Status in your symbology)
3) Scale based visibility for features so that there aren't many features on the map at higher scales
4) Checking the server settings (pooling, etc)
5) Checking the server sizing. Making sure it's properly resourced.

We have seen some improvements with using subtype group layers in our WebMaps.

Are you able to share more about what version of Enterprise you are using and possible a video of the map in action?

AmberY

Hello @gis_KIWI4 , thanks for the detailed suggestion.

1) We only symbolised the feature after published the UN with no symbol and share it as web map. Like what the community has suggested, and our symbology is mostly based on asset type which is not too complex.

2) Will definitely try on that!

3) Yes, it has been set.

4 &5) This is the tricky part for us. All these while we are on a thought that the optimum instance would be number of server core + 1. So below is what we have set. Can u spot something that we can work on?

We are using AE11.3.

Thanks again!

gis_KIWI4

The numbers and the logic looks sensible but there are too many variables in these kind of problems to find a smoking gun. It could be the number of users, you might need to check SQL logs to see if there are queries slowing down the performance. The problem could exist anywhere from the webadapter -> server -> DB

@RobertKrisher has linked the reference architecture and it would be good to compare your setup against it.

One other thing to try is tweaking values on the Parameter on the server.

Do you run DB maintenance scripts to recreate spatial indexes? This can increase performance if you do a lot of spatial changes in the data or you do sizable data loads regularly.

It can be quite frustrating not having a quick solution but work your way through the possibilities.

RobertKrisher

Make sure to familiarize yourself with the design consideration for a Network Information Management System: Reference Architecture for a Utility's Network Information Management System | ArcGIS Architecture C....

If this service is used by the majority of your editors, you should increase the minimum number of instances. Otherwise users will experience long wait times every time a new instance is started.

RobertKrisher

You can find an example test study that discusses the number of instances to consider here: Test Study for a Network Information Management System Electric Utility (PostgreSQL) | ArcGIS Archit...

RobertKrisher

@AmberY also, can you please provide repro steps (and ideally log a case with support) that shows how you're getting a 999999 error when enabling cluster keys? You shouldn't be getting an error.