R
RunPod•3w ago
ozzie

Delete Serverless Endpoint via the API?

I am trying to delete the serverless endpoint via the API, but everytime I make a request to the endpoint, I get an internal error: Via the Python API: 
delete_endpoint_graphql = """mutation {{
  deleteEndpoint(id: "{ENDPOINT_ID}") {{
    id
  }}
}}
"""


async def delete_endpoint(endpoint_id: str):
    try:
        runpod.api.graphql.run_graphql_query(
            delete_endpoint_graphql.format(ENDPOINT_ID=endpoint_id)
        )
    except runpod.error.QueryError as e:
        print(e)
        print_warning(
            f"Failed to delete endpoint {endpoint_id}: {e}Please delete it manually."
        )
delete_endpoint_graphql = """mutation {{
  deleteEndpoint(id: "{ENDPOINT_ID}") {{
    id
  }}
}}
"""


async def delete_endpoint(endpoint_id: str):
    try:
        runpod.api.graphql.run_graphql_query(
            delete_endpoint_graphql.format(ENDPOINT_ID=endpoint_id)
        )
    except runpod.error.QueryError as e:
        print(e)
        print_warning(
            f"Failed to delete endpoint {endpoint_id}: {e}Please delete it manually."
        )
Response: 
Something went wrong. Please try again later or contact support.
Something went wrong. Please try again later or contact support.
I confirmed that my min and max is set to zero as per the GraphQL docs: 
min and max workers must both be set to zero for your call to work.
min and max workers must both be set to zero for your call to work.
Solution:
Does it work when you use the REST API? https://rest.runpod.io/v1/docs#tag/endpoints/DELETE/endpoints/{endpointId}
Jump to solution
5 Replies
Solution
Dj
Dj•3w ago
ozzie
ozzieOP•3w ago
Hm yes, that worked 
async def delete_endpoint(endpoint_id: str):
    try:
        url = f"https://rest.runpod.io/v1/endpoints/{endpoint_id}"
        headers = {"Authorization": f"Bearer {os.getenv('RUNPOD_API_KEY')}"}

        response = requests.delete(url, headers=headers)
        if response.status_code != 200:
            raise Exception(
                f"Failed to delete endpoint {endpoint_id}: {response.json()}"
            )
    except Exception as e:
        print(e)
        print_warning(
            f"Failed to delete endpoint {endpoint_id}: {e}Please delete it manually."
        )
async def delete_endpoint(endpoint_id: str):
    try:
        url = f"https://rest.runpod.io/v1/endpoints/{endpoint_id}"
        headers = {"Authorization": f"Bearer {os.getenv('RUNPOD_API_KEY')}"}

        response = requests.delete(url, headers=headers)
        if response.status_code != 200:
            raise Exception(
                f"Failed to delete endpoint {endpoint_id}: {response.json()}"
            )
    except Exception as e:
        print(e)
        print_warning(
            f"Failed to delete endpoint {endpoint_id}: {e}Please delete it manually."
        )
Any reason why the REST endpoint worked? fwiw, I also didn't scale down my worker before deleting (I am able to call that endpoint with two active workers). This is fine for my usage case, I just wanted to surface in case that's unexpected behavior
Dj
Dj•3w ago
I don't know very much about the GQL API 😄 Maybe @nathaniel can explain how it's different?
nathaniel
nathaniel•3w ago
rest api just goes and calls gql for this operation so if it's allowed in rest it's also allowed in gql. I believe it actually used to be an error but was just adding another annoying step to the UI flow, so we made delete endpoint also take care of that in gql as for why the code for gql was not working before, idk :shrug: but since rest api is just doing the same thing on the backend it was probably an error in calling it. those opaque error messages from gql are unfortunate and someday we should figure out how to expose more info to users in them without revealing a bunch of internals. As devs, they are annoying to us too, but at least we can go look at the logs to see what actually went wrong the python sdk was created long before rest api existed, which is why it hits graphql all over the place
ozzie
ozzieOP•3w ago
Got it, thank you for all the additional context. FWIW, I just retried the original sample and it is still experiencing the "Something went wrong. Please try again later or contact support" error so the failure isn't intermittent. Let me know if there is any additional details that would be helpful for debugging But anyways, the REST endpoint solves my problem (automating inference evaluation runs programatically) so I'll move forward using that API moving forward. Thanks yall!

Did you find this page helpful?