Changefeeds are not distributed evenly over TiCDC nodes
Nobody has claimed this yet.
- Dominant language
- Go
- Stars
- 464
- Forks
- 313
- Avg merge
- 4d 7h
- Merged PRs (30d)
- 14
Description
What did you do?
Creating multiple changefeeds where each one contains only one table results in all table tasks being sent to a single capture. This is unreasonable. I think before we bind a table to a capture, we should add more judgment conditions, such as considering the current tasks already bound to it. Then, we can choose the best capture to bind the table to, which will be better for this scenario.
https://github.com/pingcap/tiflow/blob/master/cdc/scheduler/internal/v3/scheduler/scheduler_basic.go#L81-L82
What did you expect to see?
The tasks evenly distirbuted on different ticdc nodes.
What did you see instead?
All of them are bound on specific node.
Versions of the cluster
TiCDC version (execute cdc version):
v7.1.1
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Start with cdc/scheduler/internal/v3/scheduler/scheduler_basic.go at lines 81-82 and trace how table tasks are selected and bound to captures. Compare the current task assignments across multiple single-table changefeeds; done means tasks are distributed more evenly across available TiCDC nodes without losing correct binding behavior.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- go
- Domain
- backend, distributed-systems
- Issue type
- Bug
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Stale
- Clarity
- Mostly clear
- Newbie friendliness
- 35/100