Alex Chi Z.

icon indicating a website link https://skyzh.dev icon indicating an email [email protected]

icon company @neondatabase icon location Pittsburgh, PA, USA ⇌ Shanghai, China icon location building database systems @neondatabase, previously @cmu-db; a bubble tea every day makes me happy right away

Results 309 comments of Alex Chi Z.

Description of slow view query

Also, would you please upgrade to latest version? We've fixed some OOM issues yesterday.

Description of slow view query

By the way, when are data ingested into Kafka? After all MVs created, or before MV creation?

storage: high memory usage of compactor

The ideas look good to me. cc @hzxa21

storage: high memory usage of compactor

> S3 sdk requires a ByteStream implementing object, not necessarily Bytes like we provide now. So if we change the object store interface to accept some sort of byte stream...

Replace serde-protobuf and rust-protobuf with prost?

I believe the dependency can be removed as long as we go through the protobuf decoding thing in connectors. cc @tabVersion may take a look if you have time :)

Replace serde-protobuf and rust-protobuf with prost?

There's https://docs.rs/prost-reflect/latest/prost_reflect/index.html, probably someone is working on this.

chore(dev): use playground config for dev

Doesn’t make sense because only meta implements idle timeout. We’ll need to implement idle timeout for other components (or let risedev kill existing components) to make everything work.

batch: Add metrics for tasks.

Metrics are stored in Prometheus server, but they are not sent. Prometheus will pull metrics from each compute node.

batch: Add metrics for tasks.

What's the difference between metrics in CN and metrics in Prometheus?

feat(executor): Stateful operators should stream newly arriving messages

> > Same question. Can this be solved by turning the checkpoint interval larger? > > How to do this? In `risingwave.toml` we have a checkpoint interval config. Simply update...