Description of problem:

Every 15 minutes in /var/log/messages we log the following against each OSD residing on the OSD node.

Mar  8 08:39:33 cephosd1 kernel: libceph: osdxx 10.0.0.0:6820 socket closed (con state OPEN)

These messages are output by one of the kernel clients and report that the socket connections to the corresponding OSD have died from inactivity. Seeing these messages in an inactive or extremely lightly used cluster would be expected. These messages would also occur on OSD's which are not used regularly compared to other OSD's in the cluster.
Can these messages be outputed somewhere other than /var/log/messages as they can fill the log with useless messages and deter more of the useful messages.

Upstream Mailing list documents this issue:

http://lists.opennebula.org/pipermail/ceph-users-ceph.com/2014-August/042438.html


Version:
Red Hat Ceph 1.3.1 and Red Hat Ceph 1.3.2