Increase nginx timeouts from 10s to 30s
Testing in certain network environments has show the 10s
timeout is too small. It's unclear if the issue is with the
client or server side. When it hits, it occurs during the
upload of BMC and PNOR images.
The error on the client side is a errno 104 (Connection reset
by peer). The error on the server side is a 408 which in nginx
indicates a timeout for the client_body_timeout setting or the
client_header_timeout setting. It's doubtful the client_header
could be the issue here so only the client_body_timeout was
focused on.
The send_timeout value can also result in a dropped client connection.
The default for both the client_body_timeout and the send_timeout is
60s. In openbmc these were changed to 10s to match with OWASP
guidelines. This commit bumps them up to 30s to address this issue.
Change-Id: I7e360ed03fc533e5ca77f829859a82737fe9d5bf
Signed-off-by: Andrew Geissler <geissonator@yahoo.com>
diff --git a/meta-openbmc-machines/meta-openpower/meta-ibm/recipes-httpd/nginx/files/nginx.conf b/meta-openbmc-machines/meta-openpower/meta-ibm/recipes-httpd/nginx/files/nginx.conf
index c945544..8cf8c0d 100644
--- a/meta-openbmc-machines/meta-openpower/meta-ibm/recipes-httpd/nginx/files/nginx.conf
+++ b/meta-openbmc-machines/meta-openpower/meta-ibm/recipes-httpd/nginx/files/nginx.conf
@@ -29,10 +29,10 @@
# Comment out to enable access log in /var/log/nginx/
access_log off;
- client_body_timeout 10;
+ client_body_timeout 30;
client_header_timeout 10;
keepalive_timeout 5 5;
- send_timeout 10;
+ send_timeout 30;
# Do not return nginx version to clients
server_tokens off;