DevOps Notes

I have a strange attachment to Nagios web interface. Even though we kicked Nagios itself out of our infrastructure some time ago, we are still keeping a more moden fork of Nagios, Icinga, around as a cornerstone of our event-monitoring infrastructure. With quite a bit of effort spent writing cookbooks and monitors, Icinga is fully automated, and I can enjoy watching and showing my Nagios Web.

Nagios is loved and hated in the industry. Nagios has a great brand recognition, but suffers from less than intuitive setup and difficult to automate configuration. With passive checks and right Chef cookbooks, Nagios/Icinga setup becomes very effective.

Another Nagios weakness is poor implementation. Fortunately, Nagios poorly implements right ideas, so not everything is lost. This war story is about fixing Nagios NSCA implementation. Nagios NSCA is a Nagios passive-check daemon written in C that runs on a Nagios server and processes all passive check sent by Nagios clients.

This story is from a year ago. I kept telling myself that I got to write a blog post about it and here it is: a continuous deployment troubleshooting case.

One of the application server pools running a LAMP application (Linux / Apache/ MySQL / Memcached / PHP) was showing the following issue. Every time new code was deployed to the pool, Apache processes on 10-15 random servers out of 300+ would deadlock and stop serving requests. The problem was happening for some time but was not perceived as a big issue. Random Apache deadlocks did not impact the end users, and restarts of deadlocked Apache instances returned services to the operational state. I was not very comfortable with this behavior, so I decided to dig in and see what was going on.

Problem

Requests to a LAMP-based Facebook application are load balanced between server nodes by an F5 BIG-IP Local Traffic Manager (LTM). The F5 BIG-IP LTM has two components: an ASIC-based fast-switching component and an AMD Opteron-based software-switching component. Layer 4 load balancing between the application server nodes can be handled either exclusively by the F5 ASIC (Performance L4 mode), or exclusively by F5 software components (Standard mode), or by a combination of F5 ASIC and F5 software components. With the Standard mode turned on, F5 BIG-IP LTM capacity utilization is at 60%. With the Performance L4 mode turned on, F5 BIG-IP LTM capacity utilization is at 20%. However, with the Performance L4 mode turned on, developers report seeing an occasional blank Facebook canvas page returned in response to legitimate application requests. Serving blank Facebook canvas responses at any significant rate is unacceptable. The issue with the blank Facebook canvas responses in Performance L4 mode has to be addressed.