search

prometheus-community / elasticsearch_exporter

Elasticsearch stats exporter for Prometheus
Apache License 2.0
1.93k stars 791 forks source link

more shard-related metrics #156

Open caarlos0 opened 6 years ago

caarlos0 commented 6 years ago

It would be nice if we could have metrics like:

I'm pretty new to ES, but where it seems important to track the shard size to decide if it is a good idea to reindex with more shards for example.

The number of replicas for a given index may be important to alert if there are too few replicas - which could mean that a data loss may happen.

If you agree this information is valuable, I can work on PR :)

Just wanted to check first...

Thanks again for the great project BTW 🍻

zwopir commented 6 years ago

Hi Carlos,

yes, it sounds reasonable to track the shard metrics, but not critical. The elasticsearch_exporter exports the cluster status (red, yellow, green) which includes the information about "under-replicated" shards.

So if the metrics are already in the JSON we get from querying ES, I'm happy to accept your PR. If not or if we had to add new parameters to the curl, thus increasing the size of the transferred JSON or adding new curls to the ES, let me think twice.

Thanks for offering on the PR, I'll happily review it!

charles-mims-sp commented 5 years ago

Shard size is what I came here looking for. When shards get around 50GB or so, performance starts to really suffer.

elwood218 commented 3 years ago

Hi there. Yes it would also be nice if we could habe the number_of_shards setting per index. This can be used for calculating things. Furthermore I would be interested in that Merge what is mentioned here about shards_store_size_in_bytes. But there is no Tag and also no Docker Image build of that which included this new metric or am I missing something?!

caarlos0 commented 3 years ago

You can use caarlos0/elasticsearch_exporter:1.2.0-alpha11, it has more things though: https://github.com/justwatchcom/elasticsearch_exporter/compare/master...caarlos0:dev?expand=1