308 Permanent Redirect

hi guys! Is someone here that could help me with one small issue ?

I installed qdrant to kubernetes services, it deployed just fine. It is exposed as service in ingress route.

qdrant is accessible thought REST API (web or insomnia) but when I am using python-client it fails. Sending stack trace.
Thanks!

File ~/miniforge3/lib/python3.9/site-packages/qdrant_client/qdrant_client.py:1332, in QdrantClient.recreate_collection(self, collection_name, vectors_config, shard_number, replication_factor, write_consistency_factor, on_disk_payload, hnsw_config, optimizers_config, wal_config, timeout)
   1329 if isinstance(wal_config, grpc.WalConfigDiff):
   1330     wal_config = GrpcToRest.convert_wal_config_diff(wal_config)
-> 1332 self.delete_collection(collection_name)
   1334 create_collection_request = rest.CreateCollection(
   1335     vectors=vectors_config,
   1336     shard_number=shard_number,
   (...)
   1342     wal_config=wal_config
   1343 )
   1345 self.http.collections_api.create_collection(
...
---> 91 raise UnexpectedResponse.for_response(response)

UnexpectedResponse: Unexpected Response: 308 (Permanent Redirect)
Raw response content:
b'\r\n308 Permanent Redirect\r\n\r\n<h1>308 Permanent Redirect</h1>\r\n<hr />nginx\r\n\r\n\r\n'

Hi! can you try to open the list in browser?

from the first glance the router in k8s is misconfigured or something like this

Hi! I created a collection through insomnia and when I call in the route/collections/{colName} it works fine.

Browser works too.

hm, only delete collection doesn't?

no. Whole communication from python-client doesnt work.
just to be sure, I initialised python-client like this.

client = QdrantClient(host="kubernetes/qdrant", port=6333)

kubernetes/qdrant - that doesn't look like a valid host

well. I did not wanna paste our kubernetes address here 🤣 so I used "kubernetes" as placeholder.

i mean, /qdrart should not be there

it expects host, not url

so we need to deploy qdrant to "/" in kubernetes? Because we deployed it "/qdrant"

yeah, i guess so

When I call python-client with only host without "/qdrant" I get time out error

you can do subdomain

qdrant.kubernetes

hmm. But why web api calls with "/qdrant" works?

i think, that’s because of the client logic. it tries to construct url from /, assuming that the host is just host

give me sec. I try move it to "/"

well I tried "/" and "qdrant.kubernetes/" and both failed 😦

could you try just qdrant.kubernetes?

assuming that http://qdrant.kubernetes/ opens in browser with version info

yes

{
    "title": "qdrant - vector search engine",
    "version": "0.10.5"
}

but still same error with python-client. Web works fine

hm, interesting, you are using an old version. What's your client version?

client version

qdrant-client==0.11.5

and qdrant version I am using is the default from when I installed it with helm install as per instructions on the web site.

could you please try to change the version in your helm chart to v0.11.5

here https://github.com/qdrant/qdrant-helm/blob/main/values.yaml#L6

yeah. give me sec

we updated to v0.11.5. Web and Insomnia calls works fine but when we are trying to access from python-client we get this error.

File ~/miniforge3/envs/winehunter/lib/python3.9/site-packages/httpcore/_exceptions.py:10, in map_exceptions(map)
      9 try:
---> 10     yield
     11 except Exception as exc:  # noqa: PIE786

File ~/miniforge3/envs/winehunter/lib/python3.9/site-packages/httpcore/backends/sync.py:94, in SyncBackend.connect_tcp(self, host, port, timeout, local_address)
     93 with map_exceptions(exc_map):
---> 94     sock = socket.create_connection(
     95         address, timeout, source_address=source_address
     96     )
     97 return SyncStream(sock)

File ~/miniforge3/envs/winehunter/lib/python3.9/socket.py:844, in create_connection(address, timeout, source_address)
    843 try:
--> 844     raise err
    845 finally:
    846     # Break explicitly a reference cycle

File ~/miniforge3/envs/winehunter/lib/python3.9/socket.py:832, in create_connection(address, timeout, source_address)
    831     sock.bind(source_address)
--> 832 sock.connect(sa)
    833 # Break explicitly a reference cycle
...
     96 except Exception as e:
---> 97     raise ResponseHandlingException(e)
     98 return response

ResponseHandlingException: timed out

hmm, does it work with curl to query something except for GET?

like curl -X DELETE qdrant.k8s/collections/test_collection?

yeah this error.

308 Permanent Redirect

<h1>308 Permanent Redirect</h1>
<hr />nginx

might it be your nginx config?

I updated this line in ingress config - ingress.kubernetes.io/ssl-redirect: "false"
curl works fine now but client issue are still present.

are you using https?

yes we are]

try QdrantClient(..., https=True)

still getting same error - ResponseHandlingException: timed out

hmm, and in curl you are using curl -X DELETE qdrant.k8s:6333/collections/test_collection or just qdrant.k8s?

just qdrant.k8s

so you are not using default port?

on qdrant we set up default port 6333

try QdrantClient(..., https=True, port=443)

well, it looks like your ingress overwrote it, cause curl works without the port specified

hi @Shaibal sure, could you please create another thread?

@propher mopher did it work?

yeah! 443 works. Thank you very much. Can I have one more question?

glad to hear that. sure

i tried upsert 12k x 169 vectors
with

client.upsert(

and it raised error Unexpected Response: 413 (Request Entity Too Large) do you know where I need to set size for qdrant in k8s? Because on local host it worked without issues

@propher mopher I think the reason for that is nginx

http://nginx.org/en/docs/http/ngxhttpcoremodule.html#clientmaxbodysize

try increasing the clientmaxbody_size

or upload in smaller batches

yes. this is parameter that I was looking for in ingress

nginx.ingress.kubernetes.io/proxy-body-size: 8m

Thanks guys for help. You are the best.