We are running our site on a RHEL server. Nginx is setup to proxy
to Tomcat. We are hearing reports from some users of issues uploading
files. These issues are sporadically happening and it sounds like users
with flakey connections are the ones affected. Based on whatâ€™s
think there is an issue with a users connection timing out during the
process. This is causing uploads to fail instead of keeping the session
open until the timeout clears. The app (Grails) and Tomcat have a
long timeout set (30 min) so it doesnâ€™t appear to be something there.
wondering if there are backend timeout issues with the proxy. Maybe a
problem where Nginx closes the connection to the user or Tomcat once the
user has timed out for a period of time. Weâ€™re running Nginx 0.7.64 and
Iâ€™ve added both configuration files to pastie.org.
nginx.conf = http://www.pastie.org/908712
virtual.conf = http://www.pastie.org/908714
Iâ€™ve looked into the following directives and implementing the â€œsuggestâ€
values on staging tonight.
Default: proxy_connect_timeout 60;
suggest: proxy_connect_timeout 240;
default: proxy_ignore_client_abort off
suggest: proxy_ignore_client_abort on
default: proxy_max_temp_file_size 1G;
suggested: proxy_max_temp_file_size 4G;
default: proxy_read_timeout 60
suggested: proxy_read_timeout 240
default: proxy_send_timeout 60
suggested: proxy_send_timeout 240
suggested: client_body_timeout 240
a comment - 07/Apr/10
04:05 PM I’m going to add those configurations above and below to stage
test to make sure everything is fine. If it is, then I’ll setup
to run with these new values. client_body_timeout default: 60 suggested:
The problem with this issue is I havenâ€™t been able to reproduce it, nor
clear cut steps to reproduce. No precise timestamps have been given so
makes using the logs pretty difficult. Here is an interesting clip from
error.log that might help identify the issue. Most of my research has
this panic happens due to perl but this Nginx was compiled without the
panic: MUTEX_LOCK (22) [op.c:352].
2010/01/11 09:52:11 [error] 2250#0: *10 upstream timed out (110:
timed out) while reading response header from upstream, client:
22.214.171.124, server: fileblaze.net, request: “POST
/soundblaze/upload/fileUpload/ HTTP/1.1”, upstream: "
http://127.0.0.1:8080/soundblaze/upload/fileUpload/", host: "
Do you think those settings will help? Do I need all of these settings?
Any other suggestions I should try out? Please let me know if there is
more data you would like me to gather to help troubleshoot this.
Thanks in advance for all your help,