Out-of-memory (OOM) and out-of-threads (OOT) events and alerting
To set up out-of-memory (OOM) and out-of-threads (OOT) events for standalone/PaaS scenarios and cloud-native Full-Stack injections, follow the instructions below.
Enable the OneAgent feature
To enable out-of-memory (OOM) and out-of-threads (OOT) detection
- Go to Settings and select Preferences > OneAgent features.
- Find and turn on Enable Out-Of-Memory and Out-Of-Thread Detection for Kubernetes and PaaS installations.
- Select Save changes.
Set up high GC activity alerts
If you've already set up alerts for high GC activity in your environment, alerts are automatically created for standalone/PaaS scenarios and cloud-native Full-Stack injections.
To verify your setup
-
Go to Settings > Anomaly detection and select Hosts.
-
Make sure that Detect high GC activity is turned on.
If you're using a customized setup for long garbage-collection times alerts, note that, for standalone/PaaS scenarios and cloud-native Full-Stack injections, data collected in 10-second observation intervals is adjusted to one-minute observation intervals.
Alternatively, you can create alerts only for specific metric events.
-
Go to Settings > Anomaly detection and select Metric events.
-
Select Add metric event.
-
Define the following two events.
Metric event Metric key Threshold Violating samples Sliding window Dealerting samples High GC suspension time
builtin:tech.jvm.memory.gc.suspensionTime
25 %
3
5
4
High GC total collection time
builtin:tech.jvm.memory.gc.collectionTime
24 s
-
Select Save changes.