criteo / biggraphite

Simple Scalable Time Series Database
Apache License 2.0
130 stars 36 forks source link

Cassandra using LUCENE index plugin deployment to nodes #536

Closed zerosoul13 closed 3 years ago

zerosoul13 commented 5 years ago

Hello everyone,

I've been reading about using SASI or LUCENE indexes on Cassandra to improve performance once you are over 10M metrics (our current case).

For correctness in our setup, I would like to check with you guys if the plugin should be deployed to data and metadata Cassandra nodes alike when running in split mode or if this would only apply only for metadata nodes.

Thank in advance!

zerosoul13 commented 5 years ago

We have it on all nodes, and see a degradation of performance, compared to when not using the split and LUCENE.

geobeau commented 3 years ago

We don't use LUCENE internally. We are using SASI and have a patch for it https://github.com/criteo-forks/cassandra/commit/d7f01cb732bb9365b47d868142955cbad70dca03

The change should only be applied to the metadata cluster. The data one don't use secondary indexing