Hi! does Questdb support distributed timestamp sharding/partitioning? Instead of partitioning on a single node, can i partition over multiple nodes? similar to timescale distributed hypertable.

Hi @Michael, QuestDB cannot run in cluster mode yet. This is work in progress.

Okay. Thank you Imre.

Hello @Imre, Just out of curiosity ..... In your roadmap document I can see "High Availability - Distributed reads". Is this what you are referring to as "cluster mode" ? If yes then can I be running a distributed questdb within this year 2023 ?

Hi @Prashanthb, high availability is one of the high priority features we are actively working on and expected to be released in 2023.

Ok thanks for the info, Right now I am running multiple machines running different instances of questdb. I hope I can integrate all of them under one access node.

If your QuestDB instances have the same data, you will be able to setup replication. This way you will ingest into a single instance only and replicas will be available for backup and queries.

Just one more question @Imre 🙂 From what I am guessing, your HA design is increasing multiple query throughput but not single query latency. I had actually hoped to split the database onto multiple machines so that query processing can be parallelized, thus lowering latency. In your design of HA are you parallelizing query processing ? Just like timescaledb does ?

The aim is to create a fully distributed database where the data is sharded/partitioned and replicated across multiple nodes. This requires distributed reads, so as you hoped queries will run distributed on the cluster. However, the desired final state will be achieved in multiple steps. We have added WAL, next step is to build replication. Replicated nodes can be used as a backup and for queries (non-distributed). After replication we can start working on the distributed features which eventually can provide real clustering.

Ah ! Thanks sir, that clears all my questions. I was thinking of writing a script to distribute my database over multiple machines. I will go ahead and do that now. When you present your version I will switch over to that. Thanks. Bye.