Description of problem: in certain conditions, a logshifter process configured for syslog stops sending logs. It reads the pipe yet no messages are received on syslog side. This causes major issues with gear logging as the gear needs a restart to fix this Version-Release number of selected component (if applicable): openshift-origin-logshifter-1.8.1.0-1.el6op How reproducible: sometimes with gears having high log volume Steps to Reproduce: no clear reproducer at the moment. The gear can run for months before this issue reoccurs.
To continue diagnosing this, we'll need either: 1. Steps to reproduce. 2. A stack dump from logshifter while it's observed to be in a stalled state. This can be obtained by issuing a SIGABRT to the logshifter process.
Investigation of this bug revealed that the root cause is not in logshifter. It appears to be a problem in the kernel's epoll facility. This is being tracked in a separate bug report for the kernel (bug 1284900). The resolution of the kernel bug should address the problem reported in here for logshifter, so I'm closing this one as a duplicate. *** This bug has been marked as a duplicate of bug 1284900 ***