Possible bug with cached redirects and chunked encoding

Hi all,

I have a problem using nginx as reverse proxy cache (in front of Jetty).

Depending on the client, a redirect from cache takes either very long to
complete or fails (curl: (18) transfer closed with outstanding read data
remaining) - only chrome works as expected.

It looks like a redirect from cache doesn’t play nicely with
“Transfer-Encoding: chunked”. Using Wireshark, I’ve seen that no empty
chunk follows the header after a cache hit. I’d bet that this is illegal
(and handled differently in different clients - hence the different
behavior).

Is this a known issue? Or is there a patch available? As a quick
workaround, I’ve disabled caching of redirects.

some config snippets:

1st snippet

server {

snip

proxy_cache myapp;
proxy_cache_valid 200 302 30m;
proxy_cache_valid 301 1h;
proxy_cache_use_stale updating;
proxy_cache_key “$scheme$host$request_uri”;
add_header X-Upstream-Cache-Status $upstream_cache_status;
}

2nd snippet

location / {
# snip
proxy_pass http://backend;
proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
proxy_set_header Host $http_host;
proxy_pass_header Set-Cookie;
}

3rd snippet

http {

snip

gzip on;
gzip_disable msie6;

proxy_cache_path /tmp/nginx-cache levels=1:2 keys_zone=myapp:10m
inactive=60m max_size=1000m;
}

First request to a cached resource:

$ curl -i http://local.example.com/script/all.js
HTTP/1.1 301 Moved Permanently
Server: nginx
Date: Wed, 13 Jan 2010 16:13:57 GMT
Transfer-Encoding: chunked
Connection: keep-alive
Location: http://www.local.example.com/script/all.js
X-Upstream-Cache-Status: MISS

Second request to a cached resource:

$ curl -i http://local.example.com/script/all.js
HTTP/1.1 301 Moved Permanently
Server: nginx
Date: Wed, 13 Jan 2010 16:15:02 GMT
Transfer-Encoding: chunked
Connection: keep-alive
Location: http://www.local.example.com/script/all.js
X-Upstream-Cache-Status: HIT

curl: (18) transfer closed with outstanding read data remaining

Request to backend server:

$ curl -i http://local.example.com:8082/script/all.js
HTTP/1.1 301 Moved Permanently
Date: Wed, 13 Jan 2010 16:33:15 GMT
Location: http://www.local.example.com/script/all.js
Content-Length: 0

Cheers, Stefan

Posted at Nginx Forum:
http://forum.nginx.org/read.php?2,40752,40752#msg-40752

Hello!

On Wed, Jan 13, 2010 at 12:05:24PM -0500, sfussenegger wrote:

I have a problem using nginx as reverse proxy cache (in front of Jetty).

Depending on the client, a redirect from cache takes either very long to complete or fails (curl: (18) transfer closed with outstanding read data remaining) - only chrome works as expected.

It looks like a redirect from cache doesn’t play nicely with “Transfer-Encoding: chunked”. Using Wireshark, I’ve seen that no empty chunk follows the header after a cache hit. I’d bet that this is illegal (and handled differently in different clients - hence the different behavior).

Is this a known issue? Or is there a patch available? As a quick workaround, I’ve disabled caching of redirects.

[…]

Request to backend server:

$ curl -i http://local.example.com:8082/script/all.js
HTTP/1.1 301 Moved Permanently
Date: Wed, 13 Jan 2010 16:33:15 GMT
Location: http://www.local.example.com/script/all.js
Content-Length: 0

Looks like this problem:

http://nginx.org/pipermail/nginx/2009-December/017792.html
http://nginx.org/pipermail/nginx/2009-October/015789.html

Maxim D.

This forum is not affiliated to the Ruby language, Ruby on Rails framework, nor any Ruby applications discussed here.

| Privacy Policy | Terms of Service | Remote Ruby Jobs