Over the weekend we bounce our indexers and we just found out that the data model accelerations take over an hour to stabilize after such bounces. Their cpu is close to 100% for a while, the time to complete the searches is very long and we don’t fully trust the system when the cpu is so high for quite a long time. Any thoughts how to improve the situation?
I recommend you consider monitoring the health status logs to see if there are any recommendations present there: SPLUNK_HOME/var/log/splunk/health.log. You can also manually trigger a Health Status Check if need be. It can help identify conditions when inadequate resources are available.