I have a cluster consisting of 3 nodes of Neo4j 3.5. We are not in the production yet so for the vast majority of the time this cluster does nothing i.e. there are no writes and barely any reads.
Despite this fact when I look at JVM Heap Size I observe something suspicious (for each node):
- The heap size increases until it reaches the limit.
- Then it drops almost to 0.
- Then it increases until it reaches the limit.
- Then it drops almost to 0.
- ... and so on.
In one case, the cycle takes more than 1 hour. I also checked another cluster and the situation is similar but with some differences: the cycle is much shorter and the heap size does not drop to 0.
My questions are:
- Is it a normal situation? Should I worry?
- If it is normal, what is the reason behind that. What does Neo4j actually do?
- Why do sometimes the cycle described above takes X minutes and in another case Y minutes?
Regards,
Michał Komorowski