Cluster health dimensions: Fragment and node
The definitions of clusters, indexes, fragments, and copies are not described here. The number of fragments affects the cluster performance. If the number of fragments is too large or too low, some problems may occur.
If the number of fragments is too large, the batch write/query request is split into too many subwrites/queries. As a result, the write and query rejection rates of the index increase.
For indexes with a large amount of data, when the number of fragments is too small, node resources cannot be fully utilized. As a result, the machine resource usage is low or unbalanced, affecting the write/query efficiency.
When the GET _cluster/health is used to monitor a cluster, you can query information about the cluster status, number of nodes, and number of active segments. You can also view the number of re-located fragments, initialized fragments, and unallocated fragments.
Key indicators for cluster running:
Status: Status of a status cluster. Red: Some primary fragments are not allocated. Yellow: Partial replica shards are not allocated. Green: All fragments are allocated properly.
Nodes: Node. Includes the total number of nodes in the cluster, including the number of successful and failed nodes. Count of Active
Shards: Number of active fragments. Number of active segments in a cluster. Relocating Shards: Relocates the fragment. Number of fragments moved due to node loss.
Initializing Shards: Initialize the segment. Number of fragments initialized because indexes are added. Unassigned
Indicates the Shards. Unallocated segment. Number of fragments that have not been created or allocated.
