I had a search index on big table with primitive types of fields (text, double, integer) for indexing. For new requirements, it was decided to create new complex types for data (using "CREATE TYPE... " statements), but after it I see too high CPU consumption per node (solr joinings are disabled, but solr node on the machine this Cassandra node).
Is it correct to conclude that Solr indexing complex types blame on high CPU consumption? (Or how I can get testimonies from OpsCenter?)
@irisha_mur_186571, in addition to the questions asked by @Erick Ramirez & @alex.ott, do also post answers for the following,
As a general rule, user-defined types (UDTs) are more expensive to index than primitive CQL data types because every UDT is a nested document in Search meaning that it results in 1 additional Solr document. But how much impact it has on your cluster will vary widely as it depends on the CQL schema, Solr schema, how much nesting is involved in your UDT, etc.
If you provide additional information, I can ask one of our engineers to review it and provide their thoughts:
Please post the additional info by editing your original question. Cheers!
7 People are following this question.
