Versions (relevant - OpenSearch/Dashboard/Server OS/Browser):
Opensearch 2.12.0
Describe the issue:
We’re trying to bring down the memory use on our data nodes and discovered 2 metrics from the Prometheus exporter: opensearch_indices_segments_memory_bytes
and opensearch_index_segments_memory_bytes
. I would have assumed that the _indices_
metrics would be roughly equal to the sum of all of the _index_
metrics on a particular host, but the total of the _index_
metrics is almost 40x the _indices_
metrics. What do these 2 metrics mean?
We have quite a number of Jaeger indices that have large _index_
values type=bitset
. We only see _index_
values from our active Logstash write index.
Are the Jaeger indices being stored in a suboptimal way?