JVM crash and heap dump collection in Virtual Warehouses
Cloudera Data Warehouse automatically captures Java Virtual Machine (JVM) crash dumps and Out-of-Memory (OOM) heap dumps from Hive, Impala, and Trino Virtual Warehouse pods and includes them in diagnostic bundles for faster troubleshooting.
When a JVM running inside a Virtual Warehouse pod crashes or encounters an Out-of-Memory error,
the system automatically generates critical diagnostic files. A
The system handles these files using the following behaviors:
jvm-dump-shipper
sidecar container runs alongside each Hive, Impala, and Trino Virtual Warehouse pod. This
container monitors for and automatically forwards the following files to your central log
collection location:| File type | Trigger | What it contains |
|---|---|---|
| JVM crash
dumps <timestamp>-<hostname>-error.log |
JVM crash | Stack traces, thread information, crash analysis, and the memory state at the time of failure |
| OOM heap
dumps <timestamp>-<hostname>-heap-dump.hprof |
Out-of-Memory (OOM) error | Full heap snapshot for memory leak analysis |
- Automatic forwarding — The sidecar container automatically forwards files to the central log collection location alongside other Virtual Warehouse log files.
- Default inclusion — Generated diagnostic bundles include these files by default without requiring additional configuration.
This automated collection eliminates the need to manually retrieve crash logs from individual pods and ensures that diagnostic information is readily available when filing support cases or troubleshooting JVM-related issues.
