New Features:
-
Initial Grafana Dashboard that included sections for:
- Cluster high-level status: Cluster health charts and basic node counts
- Cluster shards: Number of cluster shards in individual states (active, initializing, relocating and unassigned)
- Thread Pools: Internal Thread Pools stats
- System: Basic Host system metrics provided as by OpenSearch
- Document and Latencies: Document indexing, deleting and searching stats
- Caches: Field data and Query caches stats
- Throttling: Indexing and merging throttling
- JVM: Basic metrics of cluster nodes JVM (Heap and GC)
-
Initial Prometheus Alerts:
- OpenSearchClusterNotHealthy: Cluster health status alert
- OpenSearchBulkRequestsRejectionJumps: High increase in the node bulk request rejections count
- OpenSearchNodeDiskWatermarkReached: High/Low disk watermark threshold has been reached on the node
- OpenSearchJVMHeapUseHigh: High JVM Heap use on the node
- OpenSearchHostSystemCPUHigh: Host system CPU load is high
- OpenSearchProcessCPUHigh: OpenSearch process CPU load is high
Known issues:
- Cluster status chart does not correctly display the cluster health (ie: Green, Yellow, Red).
- Node DiskWatermark alerts use static thresholds value while they should be using node settings value exposed in metric (see relevant issue).