Anyone seeing DocumentStore Elasticsearch performance issues?

Permanently deleted user

13. September 2019 17:13

We have two main Geocortex applications.  One has several map services, and the largest service has at most 20 layers in it.  Most of these have some fields configured for global search/instant search.  Using global search and instant search on this application usually runs very smoothly.

 

 

The second application has several map services also.  The main map service that has 70+ layers (and this is not something we can restructure).  Each layer has a few fields enabled for global search, and we are using instant search.  We routinely have issues where global search/instant search just stops responding.  If I look at Task Manager on the server, the Java process associated with the DocumentStore/elasticsearch will be running at 80%+ CPU utilization for 20-30 seconds, then drops back to near zero.  Meanwhile the web client has given up on the request.  When we start seeing these issues, often a reboot of the server will clear it up for a while.

 

 

I could allocate more memory to the Geocortex Essentials virtual server (it has 12GB available now). The configuration file for elasticsearch is configured to grab 50% of system memory (so 6GB now).  Would that really resolve this problem?

 

 

Is it possible to offload the DocumentStore core to a separate dedicated server or set up a distributed elasticsearch environment?

 

 

Has anyone else encountered performance issues like this?

0

• 

  Chris Carr-White

  26. November 2019 22:45

  Hi Carl,

   

   

  We've observed similar performance issues when Instant Search was enabled on large and complex Map Services.  In some cases, we'd seen issues where extremely large layers within a Map Service were a particular sticking point, where queries against the layer during the indexing process seemed to cause issues with the Map Service relating to ArcSOC memory availability, and could lead to this type of behaviour with the Document Store.

  What do you see in your Geocortex Core log files when this occurs, specifically relating to InstantSearch.Scanner, InstantSearch.ScannerManager, InstantSearch.Indexer, InstantSearch.IndexScheduler?

   

   

  Chris

  0

Bitte melden Sie sich an, um einen Kommentar zu hinterlassen.