Lets start the investigation step by step.
How long this spike remains?
Do you have any scheduled job e.g. backup, reporting, patching etc?
Do you have same situations with other JVM in case more than one on the same host?
Is there any antivirus running/scanning the host/application?
its take 2 hours of spike. and in some case, when the spike reach the max of threshold, the host shutdown automatically.
i dont have any scheduled job, but when im looking root cause of the spike, i found response time also get spike too.
nope. just happen in 1 host
hmm.. no more antivirus running on the host.
what should i do, for this case? and how i should solve this case?
It means something else is running on the host which is causing this situation.
Just to isolate the issue you can schedule a task for the memory snapshots.
Review and follow the instructions from the below link.
It looks like a case of regular batch activity happeneing in JVM & Heap dump analysis is the best bet. Also, have you tried drilling down to the purepaths?
I have seen in many cases the auto-sensor does a trick. You have to drill down from the point in the chart where the Heap starts increasing. Drill down to purepath, sort these purepaths by response time + SIZE and look for the purepath which are having anomaly very high RT or long tree size. Check the auto-sensors for possible indicators.