Parquet describes one file. It has nothing to say about which files make up a table, which are current, or how they change. That gap is what Iceberg fills, which is why it adopts Parquet rather than competing with it.
New on the #QuestDB blog, and where cold storage fits.
The node that uploads your cold partitions to object storage does not have to be the primary. In #QuestDB Enterprise the cold storage manager role is independent of replication, so you can assign it to a replica and keep that work off your ingestion path.
SWITCH ROLE moves the primary role between running #QuestDB Enterprise nodes. No restart. That is your planned switchover and your failover, same statement.
QWP clients hold their own host list, so they find the new primary and replay whatever was never acknowledged.
There is a #QuestDB Enterprise Kubernetes Operator now, on EKS and AKS. Pods, storage, config, scheduled backups, restore, point-in-time recovery, replication and promotion.
It holds no cloud identity and touches no bucket. The database pods do all the object-store I/O.
We just shipped #QuestDB Enterprise 4.0.
Cold storage ages old partitions out to your bucket, where they keep answering queries. Promotion is restart-free, a SQL statement away. And QWP buffers unconfirmed rows on the client, so a node going away is a pause rather than a loss.
