Skip to content

feat(starrocks): add MySQL client task and configure local cluster for CN queries#961

Open
mbrobbel wants to merge 1 commit into
sirius-db:devfrom
mbrobbel:starrocks-client
Open

feat(starrocks): add MySQL client task and configure local cluster for CN queries#961
mbrobbel wants to merge 1 commit into
sirius-db:devfrom
mbrobbel:starrocks-client

Conversation

@mbrobbel

Copy link
Copy Markdown
Member

Add a client pixi task and client environment that open a mysql CLI against the local FE's query_port, so an FE + Rust CN can be exercised with real queries from a second terminal while cluster streams logs in the first.

Configure the local cluster so the storage-less compute node can actually serve those queries:

  • Pin the FE to loopback (priority_networks = 127.0.0.1/32). Otherwise the FE picks a non-loopback interface as its self identity and advertises it to the CN over heartbeat, mismatching the CN's --fe-host 127.0.0.1 default and leaving the client unable to reach query_port at 127.0.0.1.
  • Run in shared-data mode (run_mode = shared_data, enable_load_volume_from_conf = false). In shared-nothing mode the FE schedules scans and FILES() schema inference only on BE nodes, so a CN-only cluster fails with "No alive backends". Shared-data treats the CN as the worker and needs no object storage for external file scans.

…r CN queries

Add a `client` pixi task and `client` environment that open a `mysql` CLI
against the local FE's query_port, so an FE + Rust CN can be exercised with
real queries from a second terminal while `cluster` streams logs in the first.

Configure the local cluster so the storage-less compute node can actually
serve those queries:

- Pin the FE to loopback (priority_networks = 127.0.0.1/32). Otherwise the FE
  picks a non-loopback interface as its self identity and advertises it to the
  CN over heartbeat, mismatching the CN's --fe-host 127.0.0.1 default and
  leaving the client unable to reach query_port at 127.0.0.1.
- Run in shared-data mode (run_mode = shared_data, enable_load_volume_from_conf
  = false). In shared-nothing mode the FE schedules scans and FILES() schema
  inference only on BE nodes, so a CN-only cluster fails with "No alive
  backends". Shared-data treats the CN as the worker and needs no object
  storage for external file scans.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant