I've been testing out graylog2-server and graylog2-web-interface (both version 0.10.0) for the past week or so. And it seems very nice.
Everything logged from our hosts by syslog works perfectly, but we are seing a very weird periodic error in the web interface, when one of our rails applications is logging to the graylog2-server using the gelf gem, setup by:
config.lograge.enabled = true
config.lograge.log_format = :graylog2
config.logger = GELF::Logger.new("graylog2.somehostname.org", 12201, "WAN", { :host => "toshokan_dev", :facility => "toshokan" })
When the applications logs, and the data hits the frontpage of the webapplication a reload of this page sometimes breaks with this in the logs:
ActionView::Template::Error (invalid byte sequence in UTF-8):
54: <%= message.additional_fields[col].blank? ? "-" : message.additional_fields[col] %>
55: <% end %>
56:
57: <%= message.message[0..message_length] %>
58: <%=raw (message.message.length > message_length) && !defined?(dont_show_links) ? "..." : nil %>
59:
60:
And sometimes a force reload in the browser fixes it, and sometime it doesn't.
When the data scrolls out of the frontpage, everything is fine there again, but a search for the application log will periodicly fail, in the same way.
I tried the slice fix from #113 manually, but it doesn't seem to fix the issue. I've even tried to alter:
message.message[0..message_length]
to:
message.message
And the web application still fails periodicly, when showing logs from the rails application.
The server and web application is installed on a Debian 7.0 Wheezy server, with the following debian packages:
mongodb-10gen (latest from 10gen repository)
openjdk-7-jre-headless
elasticsearch-0.20.5.deb (latest elasticsearch from elasticsearch.org
bundler
rubygems
rubygems-integration
apache2
apache2-mpm-prefork
libapache2-mod-passenger
ruby1.9.1 (is really a ruby 1.9.3.194-8 - debian naming is because om API name I think?)
and then the rest of the web application basicly setup as described in your documentation for Debian 6.
We see the same issue when starting a test server on port 3000 with:
rails server -e production
within the web application directory, so it's not the passenger+apache2 setup that messes with us.
Is this a bug, or do you have any idea what might be done wrong in my setup?
Best regards,
Lasse Aagren.
PS: Let me know if you need more information about the setup.