CODEWITHSUNDEEP
apache-kafkaconfluent-schema-registry
pacman
pacman

Reputation: 835

Concerns on using schema registry

I have a kafka cluster that uses schema files to write and read from kafka topics. I'm currently looking to use schema registry rather than using files to load schemas because the schemas the number of schemas to manage are getting out of hand.

But the current usecase is a realtime usecase scenario. And I have some concerns while trying to use schema registry. Currently all the schemas are loaded in memory and is hence fast but schema registry would be hosted separately it creates a few questions for me.

  1. Would there be an increased delay in reading and writing data to kafka topics? If there is what would be the expected increased delay? The deployment of all applications would be within a kuberntes cluster.

  2. Is schema registry made to work for realtime usecase scenarios? I couldn't find a lot of articles that talk about this?

  3. Finally is there any other concerns that I would have to worry about that cause problem for performance and scalability?

Upvotes: 1

Views: 1013

Answers (1)

OneCricketeer
OneCricketeer

Reputation: 191671

Confluent Schema Registry is used at companies handling hundred of millions events per day (disclaimer, I used to work at one), so yes, it handles real-time just fine.

There is only a delay on the initial request; the Confluent Serializers cache the schemas in addition to the caching layer on the server. The delay should only be expected to be a few milliseconds

If you are running the registry yourself, refer 17 ways to mess up self-managed Registry

Upvotes: 1

Related Questions