Troubleshooting Java Application with 31G Heap and 500G Off-heap (Using Apache Ignite as an Example)

Anyone who is familiar with highload and distributed systems knows that sooner or later incidents occur (performance degradation or loss of cluster nodes).

As an example, we'll look at the node of the distributed database — Platform V DataGrid. This DB has found its application in hundreds of Sber systems, including critical ones: the card processing system, acquiring system and for storing profiles of all individual clients.

If the system fails, it is necessary to either determine the cause of the failure or improve performance. My talk is exactly about this: how to treat a Java application with a large amount of data when something goes wrong. We'll also cover settings of Java and Linux so that incidents do not happen.