Refreshing index pattern does not work due to timeouts

Hi Open Distro team,

First of all, thank you for the awesome software and your ongoing efforts.

I have deployed two Kubernetes environments with Elasticsearch 7.3.2, one of them is using Open Distro (1.3.0 version) and the other one the official Elasticsearch. The cpu and memory resources provided are the same in both environments. Also, the environments contain the same indexes, in both cases they have been restored using snapshots.

I’m doing a performance comparison and, while some of the queries achieve similar results, others are much slower in the Open Distro environment, even getting timeouts when the number of indexes is increased. In particular, the time required to process the following request is significantly different:

time curl -X GET "localhost:5601/api/index_patterns/_fields_for_wildcard?pattern=<index_pattern_name>*&meta_fields=_source&meta_fields=_id&meta_fields=_type&meta_fields=_index&meta_fields=_score?pretty" -H 'kbn-xsrf: true'

Elasticsearch environment: 0m0.517s
Open Distro environment: 0m6.263s

It makes no difference whether it is done from the Kibana interface or terminal.

Also, the time difference is similar when performing the request from elasticsearch/odfe instead of Kibana: time curl -X GET "localhost:9200/<index_pattern_name>*/_field_caps?fields=*&pretty"

Since none of the environments are indexing alerts at that moment, the time difference does not make any sense.

I have two questions:

  1. If Open Distro is a “plugin of Elasticsearch”, why is there such a time difference in the requests?

  2. Is this a known issue? Open Distro using more resources than Elasticsearch? And if so, why does it happen? What operations can be affected?

Best regards,
Mayte Ariza.

1 Like