[BUG] Wrong execution identity when "Recovering" stopped executions

[ddl-rliu]

Describe the bug

When user A "recovers" via Flyteconsole UI an execution belonging to user B, it launches with user B's execution identity. This can have negative security implications. See "Additional context to reproduce".

Expected behavior

When user A recovers user B’s execution, it does not successfully launch with user B's execution identity. (as this can have negative security implications)

Open question: What is the expected behavior? Should the recover fail? Or should it succeed, but it is launched with user A's execution identity? (or, is this intentional?)

Additional context to reproduce

Steps to reproduce

Pre-requisite:

• Be logged in as a user A, and view an aborted execution launched by a user B

Expected behavior:

• When user A recovers user B’s execution, it does not successfully launch with user B's execution identity. (as this can have negative security implications)

Actual behavior:

1. As user A, trigger a recover of user B’s execution (This can be done in Flyteconsole UI)
2. Check and see that the execution succeeds and is launched with user B's execution identity. (See Execution Identity for details)

Execution Identity

You can use a tool like grpcui to query ListExecutions

grpcurl -plaintext -H "<user A token>" -d '{
  "id": {
    "project": "66ac1f956fd5b26a885904ce",
    "domain": "development"
  },
  "limit": 5,
  "sortBy": {
    "key": "created_at"
  }
}' localhost:9089 flyteidl.service.AdminService.ListExecutions

Then select the execution from the response, and view the execution identity at response.executions[0].spec.security_context.run_as.execution_identity

Relaunch

The above all applies to the RelaunchExecution grpc endpoint as well, however this endpoint does not appear to be exposed via the UI so it is a less immediate concern.

Relaunch via gRPC

You can use a tool like grpcurl to query RelaunchExecution

grpcurl -plaintext -H "<user A token>" -d '{
  "id": {
    "project": "66ac1f956fd5b26a885904ce",
    "domain": "development",
    "name": "easily-ruling-crappie-aeps"
  }
}' localhost:9089 flyteidl.service.AdminService.RelaunchExecution

(note that this endpoint is not exposed via e.g. the Flyte UI)

Screenshots

No response

Are you sure this issue hasn't been raised already?

• [X] Yes

Have you read the Code of Conduct?

• [X] Yes

[runllm[bot]]

The issue described involves a security concern where user A can relaunch an execution belonging to user B, and the execution runs with user B's identity. This behavior is problematic as it can lead to unauthorized access and actions.

According to the ExecutionRelaunchRequest documentation, the RelaunchExecution endpoint does not inherently handle user identity changes. The expected behavior should be that the relaunch either fails or succeeds but runs under user A's identity.

Would you like more detail on how to address this issue or the specific steps to reproduce it?

Sources:

• ExecutionRelaunchRequest documentation

---

This is an AI-generated response and your feedback is appreciated! Please leave a 👍 if this is helpful and 👎 if it is not.

Mention @runllm to ask follow-up questions. See runllm.com for more!

[ddl-rliu]

Some extra context:

Intentional behavior

I am not sure, if this bug could actually be intentional behavior. The execution spec contains, in addition to the spec.security_context.run_as.execution_identity field, a spec.metadata.principal field. Then, for example, the flyteconsole UI has a "Only my executions" toggle, which filters based on metadata.principal, not execution_identity.

More extra context: (more relevant for the "relaunch" version of this bug)

gRPC vs HTTP

The steps to reproduce applies to the corresponding http endpoint as well; that is, /api/v1/executions/relaunch for RelaunchExecution. However it is easier to work with gRPC (e.g. doesn't require manually writing protobuf payloads), in terms of reproducing the behavior manually.

Flyte UI Relaunch

Flyteconsole allows users to relaunch executions, however it uses a different endpoint /api/v1/executions for CreateExecution. In fact the /relaunch endpoint appears unused in flyteconsole. Therefore the bug must be reproduced manually (i.e. not via natural UI usage)

[ddl-rliu]

I've updated the issue with a focus on the 'Recover' operation (which is possible via UI).

In addition to "Recover", there a similar bug with the Relaunch grpc endpoint. When user A relaunches user B’s execution, it will launch with user B's execution identity. (this too can have negative security implications)

Open question here: What is the expected behavior? Should the relaunch fail? Or should it succeed, but it is launched with user A's execution identity? (or, is this intentional?)

[ddl-ebrown]

Any comments on this one @eapolinario - I believe the behavior here is still not well understood

[github-actions[bot]]

"Hello 👋, this issue has been inactive for over 90 days. To help maintain a clean and focused backlog, we'll be marking this issue as stale and will close the issue if we detect no activity in the next 7 days. Thank you for your contribution and understanding! 🙏"

[github-actions[bot]]

Hello 👋, this issue has been inactive for over 90 days and hasn't received any updates since it was marked as stale. We'll be closing this issue for now, but if you believe this issue is still relevant, please feel free to reopen it. Thank you for your contribution and understanding! 🙏