In test environment, we have one 4-core GeoEvent server 10.9.1 (Red Hat 8.5, VM), we upgrade memory to 128GB (from 32GB) and modify ArcGISGeoEvent-wrapper.cfg file with these lines; (we want to allocate 32GB for JVM heap)
wrapper.java.additional.15=-Xms32g
wrapper.java.additional.16=-Xmx32g
After starting GeoEvent Service, everything is fine, application is operational.
Then we import a geofence with 41500 features, it is successfully imported. When we check the GeoEvent java process memory it was 20GB.
First everything looks OK, but after a few minutes (no service/input/output is running) GeoEvent Manager page requests are starting to slow down (dojo requests staying in pending state)
Then we tried to stop and start GeoEvent sevice (with systemctl), GeoEvent Manager first says "Initializing, please wait" then "Unrecoverable, please contact administrator". When we checked the memory it was 3-4GB.
Reboot, service stop/start, nothing worked to run GeoEvent again. Therefore, we did administratively reset, then everything got back.
We did same steps 2-3 times, everytime the result is same.
Please let us know what are we doing wrong?
By the way, we want to set heap size much higher than the 32GB (maybe 128 or 256GB). In Esri documentation there is a limit for Windows that max heap size can be 32GB. But there is no information for Unix, does anyone know anything about it? (I know in theory for 64-bit systems its 2^64)
