Here’s the full breakdown of flow run states in Prefect 3.x. A useful mental model is: - State type = the broad category (

SCHEDULED

,

RUNNING

, etc.) - State name = the specific label you see in the UI (

Late

,

InfrastructurePending

, etc.) Some names are just special cases of the same underlying type. Docs / references: - Prefect docs - State reference - Client schema reference --- ## 1) Core state types These are the main state categories a flow run can be in: | Type | Meaning | Terminal? | |---|---|---| |

SCHEDULED

| The flow run is planned for execution, but not running yet | No | |

PENDING

| The flow run has been submitted and is waiting on prerequisites or infrastructure | No | |

RUNNING

| The flow run is actively executing | No | |

PAUSED

| Execution has stopped and is waiting to be resumed | No | |

CANCELLING

| Cancellation has been requested and cleanup is in progress | No | |

COMPLETED

| The flow run finished successfully | Yes | |

FAILED

| The flow run finished unsuccessfully due to an execution/code error | Yes | |

CRASHED

| The flow run stopped because of an infrastructure/system issue | Yes | |

CANCELLED

| The flow run was cancelled | Yes | --- ## 2) Full list of common flow state names and what they mean ### Scheduled-type states These all fall under

SCHEDULED

. -

Scheduled

The run is set to start at a future time. -

Late

The scheduled start time has already passed, but the run has not been picked up yet. Usually this means a worker hasn’t submitted it to infrastructure yet. Common causes: - no active worker polling the right work pool / queue - worker is unhealthy or offline - concurrency limits are blocking execution - infrastructure provisioning is delayed before pickup -

AwaitingRetry

The flow run failed on an attempt, but retries remain, so it is waiting for the retry delay before trying again. -

AwaitingConcurrencySlot

The run is ready in principle, but it must wait for a concurrency slot before it can proceed. -

Resuming

The flow was paused or suspended and is scheduled to resume. --- ### Pending-type states These all fall under

PENDING

. -

Pending

The flow run has been submitted but hasn’t started execution yet. It’s waiting for preconditions to be satisfied. -

Submitting

Prefect is actively submitting the run to execution infrastructure. -

InfrastructurePending

The run has been submitted, and Prefect is waiting for the underlying infrastructure to become ready. Example: a container, pod, or job is being created but isn’t running user code yet. This is the one you mentioned as infra pending. --- ### Running-type states These fall under

RUNNING

. -

Running

The flow code is currently executing. -

Retrying

A retry attempt is actively executing after a prior failure. --- ### Paused-type states These fall under

PAUSED

. -

Paused

Execution is intentionally stopped and waiting for manual or programmatic resume. -

Suspended

Similar to paused, but the process has exited and will need to be resumed later rather than just continuing in-memory. --- ### Completed-type states These fall under

COMPLETED

and are terminal. -

Completed

The flow finished successfully. -

Cached

The result was retrieved from cache instead of recomputing. -

RolledBack

The run reached completion but a transaction was rolled back and rollback hooks were executed. --- ### Failed-type states These fall under

FAILED

and are terminal. -

Failed

The run ended with an error/exception and no more retries remain. -

TimedOut

The run exceeded its configured timeout and was marked failed. --- ### Crashed-type states These fall under

CRASHED

and are terminal.

-

Crashed

The run stopped because of an infrastructure or system-level issue rather than a normal code exception. Typical examples: - process killed - pod evicted - out-of-memory - worker/interpreter interruption - abrupt infrastructure shutdown --- ### Cancelling / Cancelled -

Cancelling

(

CANCELLING

) A cancellation request has been issued, and Prefect is attempting to stop the run and clean up infrastructure. -

Cancelled

(

CANCELLED

, terminal) The run has been cancelled and will not continue. --- ## 3) Quick “what should I infer?” guide ### If you see Late Think: “This run should have started already, but nothing picked it up.” Check: - is a worker running? - is it polling the correct work pool / queue? - is the deployment assigned correctly? - are concurrency limits blocking it? ### If you see InfrastructurePending Think: “Prefect has handed this off, but the execution environment isn’t ready yet.” Check: - container/job/pod startup delays - cloud batch system delays - image pull time - cluster capacity / autoscaling ### If you see Pending Think: “The run is in the handoff phase before actual code execution.” ### If you see Crashed Think: “This looks more like environment/infrastructure failure than user code failure.” ### If you see Failed Think: “The flow code itself raised an error or otherwise failed logically.” --- ## 4) Typical lifecycle A normal run often looks like:

text
Scheduled
  -> Pending
  -> Running
  -> Completed

A delayed pickup might look like:

text
Scheduled
  -> Late
  -> Pending
  -> Running
  -> Completed

A retrying run might look like:

text
Running
  -> Failed
  -> AwaitingRetry
  -> Retrying
  -> Completed

A run waiting on infra might look like:

text
Scheduled
  -> Pending
  -> Submitting
  -> InfrastructurePending
  -> Running

--- ## 5) Terminal vs non-terminal summary Terminal states: -

COMPLETED

-

FAILED

-

CRASHED

-

CANCELLED

Non-terminal states: -

SCHEDULED

-

PENDING

-

RUNNING

-

PAUSED

-

CANCELLING

--- If you want, I can also give you: 1. a flow-state transition diagram, or 2. the same breakdown for task run states specifically.