Motivation
We have seen abrupt index errors in production applications where an index can not recover from a lost node. When the node comes up again, the index should be rebuilt. However, some indexes are stuck in an error state, which might only be detected through application health checks: some queries never return within a specific deadline.
Related tools
Couchbase allows the retrieval of index statistics. [1] A Prometheus exporter exists to convert the data to Prometheus format. [2]
Problem
It is unclear from [1, 2] whether the index’s state can be exported to Prometheus. We aim to receive a Grafana alert in case an index is malfunctioning.
Question
Can the index state (ready, building, error, etc.) be retrieved from Couchbase index statistics?
[1] Index Statistics API | Couchbase Docs
[2] couchbase-exporter/pkg/collectors/index_collector.go at master · couchbase/couchbase-exporter · GitHub