Ceph » rgw

If I exeucted the following command.

# s3cmd ws-info s3://test2
Bucket s3://test2/: Unable to receive website configuration.

the http response is captured as follows:

# ngrep -d lo port 7480 -W byline
…
##
T 10.231.252.119:7480 -> 10.231.252.119:46904 [AP] #6
HTTP/1.1 404 Not Found.
Content-Length: 233.
x-amz-request-id: tx000000000000000000003-005d789467-2e7306-default.
Accept-Ranges: bytes.
Content-Type: application/xml.
Date: Wed, 11 Sep 2019 06:30:00 GMT.
Connection: Keep-Alive.
.

##
T 10.231.252.119:7480 -> 10.231.252.119:46904 [AP] #8
<?xml version="1.0" encoding="UTF-8"?><Error><Code>NoSuchWebsiteConfiguration</Code><BucketName>test2</BucketName><RequestId>tx000000000000000000003-005d789467-2e7306-default</RequestId><HostId>2e7306-default-default</HostId></Error>
##
T 10.231.252.119:7480 -> 10.231.252.119:46904 [AP] #10
<WebsiteConfiguration xmlns="http://s3.amazonaws.com/doc/2006-03-01/"></WebsiteConfiguration>
##
…

The last part(separated by ##) of the message is generated in RGWGetBucketWebsite_ObjStore_S3::send_response method, which should not be generated in this case.
If the http request contains an HTTP_ORIGIN header, the op_ret is changed to 0 in the RGWOp :: generate_cors_headers method, and RGWGetBucketWebsite_ObjStore_S3 :: appends the success response after the failure response.
Nevertheless, in most cases, there is no problem because a http client cuts the message by a content-length header. However, sometimes when I use it with a reverse proxy, I get an bad gateway error.