How to find out when data loss occurs in an elasticsearch cluster

Question

Is there a general way in ES to know "oops, hots, html, index to be rebuilt"?

Alternatively, a template or approach that answers this question?

So far, we have used the following approach:

1) If the cluster goes red, the data is lost, the index must be rebuilt.

2) If the cluster shimmers between green and yellow, no data loss has occurred.

3) Like # 2, on an index with 5 shards, if the value of "active_shards" is equal to or greater than 5, all is well.

Is # 3 honest? Basically, this is correct:

DataLossHasOccurred == ("active_shards" <"active_primary_shards")

Answer 1

Not. Not.

3 - equivalent check to others.

When the cluster is “red”, it means that some data is not available. He cannot be lost. If multiple servers are sent offline but can be returned, data can be restored. When this happens, the cluster will return to green.

When the cluster is "yellow", it means that the cluster is working with a smaller one. Depending on the number of replicas configured for acting indexes, this may or may not be a problem. The metrics that I use for monitoring are the general condition and the number of unallocated fragments. If the cluster is in a yellow state, and the number of unallocated fragments does not decrease, something is incorrectly configured.