Rudder

It has been observed that a Rudder server with a few number of nodes attached and otherwise mostly inactive, still shows a heavy disk write activity of about 3.6 GB/hour or ~86 GB/day.

This has been narrowed down to the activity of the Rudder agent on the Rudder server.

This has been further narrowed down to a couple ncf rules in /var/rudder/cfengine-community/inputs/common/1.0/update.cf, in which commenting out the following section cuts the overall server write activity down by -85% to -90% :

    # root_server::
    #   "${g.rudder_ncf}/common" 
    #     copy_from         => copy_digest_without_perms("${g.rudder_ncf_origin_common}"),
    #     depth_search      => recurse_ignore("inf", @{g.ncf_excludedreps}),
    #     perms             => u_mog("644", "root", "0"),
    #     action            => immediate,
    #     move_obstructions => "true",
    #     classes           => success("rudder_ncf_common_updated", "rudder_ncf_common_update_error", "rudder_ncf_common_updated_ok"),
    #     comment           => "Update the common Rudder ncf instance";

    #   "${g.rudder_ncf}/local" 
    #     copy_from         => copy_digest_without_perms("${g.rudder_ncf_origin_local}"),
    #     depth_search      => recurse_ignore("inf", @{g.ncf_excludedreps}),
    #     perms             => u_mog("600", "root", "0"),
    #     action            => immediate,
    #     move_obstructions => "true",
    #     classes           => success("rudder_ncf_local_updated", "rudder_ncf_local_update_error", "rudder_ncf_local_updated_ok"),
    #     comment           => "Update the local Rudder ncf instance";