we are running CentOS 6 servers with multiple lxc-containers. For system logging we are using syslog-ng. After a while the syslog-ng daemon stops logging messages, but the daemon keeps running. This happens on the host and inside the containers (where another syslog-ng is running) as well. We could not find any patterns for the failure yet but we assume that it has something to do with lxc, because we don't have these problems on other hosts.
We have the suspicion that these problems occur when more than on lxc-container is running and that only "new" processes can not log.
We are running the following software versions:
- CentOS-Linux 6.4/6.5
- lxc-0.7.5
- syslog-ng-3.2.5
Do you have any ideas?
Best regards trademesh
My suspicion would be that there's an FD leak (you can confirm it with lsof), which, for some odd reason, surfaces only under LXC. I'd recommend upgrading to a newer syslog-ng (3.5+), and see if the problem persists. A lot of issues were fixed since 3.2.