Linux2Appd:Get hundreds of Linux statistics from proc files into Appdynamics

linux2appd gives Appdynamics incredibly detailed statistics about a Linux system. On a typical system it will produce around 500 statistics in 12 categories. The Categories are:


  • sysinfo (number of CPUs, amount of swap, etc)
  • cpustats (idle, system, user, wait, etc for each CPU core on the system)
  • procstats (number of processes running, in queue, blocked, etc)
  • memstats (amount of memory active, buffers, dirty, inactive, free, etc)
  • pgswstats (page faults, pages in, pages out, etc)
  • netstats (bytes in/out, packets in/out, dropped, errors, etc for each network device on the system)
  • sockstats (number of tcp, udp, raw, used sockets)
  • diskstats (bytes in, bytes out, major, minor for each disk device on the system)
  • diskusage (free, total, used, percentage used for each mount on the system)
  • loadavg (1,5, 15 minute load averages)
  • filestats (allocated, free, unused, etc)
  • processes (aggregates of nlwp, stime, utime, vsize, etc for all process on the system)



Two files are provided (monitor.xml and linux2appd). These files need to be put into a directory below AppDynamics/MachineAgent/monitors in your Machine Agent install. The linux2appd script requires the Perl library Sys/Statistics/Linux which is available on CPAN ( or possibly by an operating system package. Once all is in place, restart the machine agent to have it read the monitor.xml file (the machine agent log file will give some indication that it's being used correctly).


It's possible to simply run linux2appd on the command line (as any system user) - it'll print statistics to STDOUT, so it should be easy to make sure it works on your system before having to debug the machine agent config.


See your Statistics

The statistics generated are to be found in:


Analyse -> Metric Browser -> Application Infrastructure Performance -> [Application name] -> Custom Metrics -> Linux

(or: Analyse -> Metric Browser -> Application Infrastructure Performance -> [Application name] -> Individual Nodes -> [Node name] -> Custom Metrics -> Linux)


Update 2nd Jan, 2014: It looks like Appd have now thought of this themselves: (not sure if it returns as many stats, but it looks very similar)