# TiCDC Changefeed Control Plane

## Question

At which create, update, delete, list, status, or effective-configuration stage
is the changefeed blocked, and is the observed API state consistent with the
actual TiCDC task state?

## Enter When

Use for a changefeed stuck creating/updating/deleting, unexpected list/status
results, control-plane retries, quota or RCU admission, or an API value that
differs from TiCDC's effective configuration.

## First Batch

1. Identify the exact changefeed, independent TiCDC version, deployment type
   and kernel architecture when available, current task state, and relevant
   creation or update window. Wetech OP does not require Cloud Dataflow
   metadata.
2. For Cloud, align control-plane Activities and supported task metadata with
   narrow Dataflow or rep-cdc/TiCDC logs. For Wetech OP, use its bounded TiCDC
   metrics/logs and supplied task identity for the requested lifecycle
   transition.
3. Distinguish request acceptance, resource/admission scheduling, TiCDC task
   creation, and effective configuration. A Cloud API representation mismatch
   is not automatically a TiCDC runtime failure.

## Decision

Name the earliest observed lifecycle stage that did not advance. A long
`creating` state requires positive evidence before attributing it to RCU,
Regions, or TiCDC. Compare requested versus effective configuration field by
field and preserve casing or enum normalization as a control-plane boundary.

## Strongest Alternative And Stop

Check replication progress only when the changefeed reached a running state.
Stop after one blocked lifecycle stage or one exact API/effective-state mismatch
is bounded. Never retry, pause, resume, delete, or recreate the task.