Hi there, my first contribution on this project! Let me know your thoughts.

Context

Inspired by cerebro, I want to create a dashboard in Grafana to visualize shard distribution, state and movements over time. This helps during scale up periods and diagnose imbalances. I tried to use count(elasticsearch_indices_shards_docs) by (index, node) metric, but that didn't give me proper node name (e.g. 4n6k9UN4SZG-n3MjOAnszA) and a distinction of primary, replica and shard states.

Changes

• Fix: add elasticsearch_node_shards_total on metrics.md (missing)
• Refactor: simplify shards.go to be more consistent with other files like indices.go
• Feat: add elasticsearch_node_shards_state metric with the following encoding (chosen arbitrarily):

0=unassigned
10=primary started
11=primary initializing
12=primary relocating
20=replica started
21=replica initializing
22=replica relocating

This allows me to create a Matrix with nodes as rows, indices as columns, and color the cells based on the value. The metric is flexible, has more labels but still low in cardinality due to limited possible values.