Details
-
Bug
-
Status: Open
-
Major
-
Resolution: Unresolved
-
1.1.0
-
None
-
None
Description
Upload content via S3 gateway with "Content-Type: plain/text; charset=utf-8" header will fail with 500 code. At the same time, "Content-Type: plain/text" (without charset) and "Content-Type: plain/text; charset=UTF-8" (in uppercase) were accepted.
We're running the official Ozone v1.1.0 release.
These lines shown as below are known behavior:
- 'text/plain' ==> OK
- 'text/plain; charset=UTF-8' ==> OK
- 'text/plain; charset=utf-8' ==> 500
- 'text/html; charset=UTF-8' ==> OK (same as text/plain case)
- 'text/html; charset=utf-8' ==> 500 (same as text/plain case)
- 'text/markdown; charset=UTF-8' ==> OK
- 'text/markdown; charset=utf-8' ==> OK
I've attached a code snippet (upload.go) and a log from S3 gateway (s3g.log). This code can run easily with following command lines.
% go run upload.go --contenttype='text/plain; charset=utf-8'
text/plain; charset=utf-8
InternalServerError: Internal Server Error
status code: 500, request id: , host id: exit status 1
% go run upload.go --contenttype='text/plain; charset=UTF-8'
text/plain; charset=UTF-8
&\{http://example.com/<BUCKET>/<KEY> <nil> <nil>}%
HTTP request from upload.go is here:
PUT /<BUCKET>/<KEY> HTTP/1.1 Host: localhost:9999 User-Agent: aws-sdk-go/1.38.46 (go1.16.4; linux; amd64) S3Manager Content-Length: 4 Authorization: AWS4-HMAC-SHA256 Credential=ksugihara@EXAMPLE.COM/20210525/dummy/s3/aws4_request, SignedHeaders=content-length;content-md5;content-type;host;x-amz-content-sha256;x-amz-date, Signature=<HASH> Content-Md5: AzvZSxFo1+Tw1kTDyV41vw== Content-Type: text/plain; charset=utf-8 X-Amz-Content-Sha256: <HASH> X-Amz-Date: 20210525T080147Z Accept-Encoding: gzip TEST^C
Investigation
We thought the issue is similar to HDDS-4856 but this issue will happen when empty Content-Type header was given. Our facing problem will appear when Content-Type defined together with the charset directive.
The stack trace of S3 gateway indicates the request proceeded to remote invocation and failed with authentication error with org.apache.hadoop.ipc.RemoteException(org.apache.hadoop.security.token.SecretManager$InvalidToken). There are many parts to preprocess a request before reaching RPC Client, so I suspect some Java servlet filters like ozone.s3.HeaderPreprocessor and hdds.server.http.NoCacheFilter. But as far as I found out, these are not a cause of this problem directly (because these are almost for generating response header). I also suspected some components like OzoneClientProducer, Signature Processor and HeaderPreprocessor. I think this is not direct cause. How do I investigate this problem in more detail?