I have several kvm/QEMU virtual guests (hardened Gentoo), running on several different Gentoo hosts. Nullmailer-send isn't getting invoked, and therefore mail jams up in the queue, and is never sent. Observation 1: Manually invoking nullmailer-send as user 'root' within the guest VM works - the queue is flushed and mail gets delivered as-expected. Observation 2: In addition to VM guests, I have many standalone hardened servers with nullmailer installed. These are historical/legacy machines, where nullmailer has worked fine for ~15yrs, and continues to do so. So, our site-configuration is well-known, well-tested, well-documented. Observation 3: A crucial difference appears between the working/legacy nullmailer, and the more-recent non-working kvm VMs: Working-server directory structure: # ls -al /var/nullmailer/ total 1 drwxr-xr-x 5 root root 144 Jun 23 2015 . drwxr-xr-x 17 root root 496 Mar 6 2013 .. drwxrwx--- 2 nullmail nullmail 96 Jun 23 03:12 queue drwxr-xr-x 4 root root 152 Jun 23 2015 service drwxrwx--- 2 nullmail nullmail 120 Jun 23 03:12 tmp prw-rw---- 1 nullmail nullmail 0 Jun 23 03:12 trigger Non-Working-server directory structure: # ls -al /var/nullmailer/ total 0 drwxr-xr-x 3 nullmail root 20 Jun 20 12:00 . drwxr-xr-x 13 root root 154 Oct 18 2016 .. drwxr-xr-x 4 root root 46 Jun 21 11:15 service For both the working and non-working situations, nullmailer runs under svscan, and uses the openrc init-system. Kernel-notification features are the same between working/non-working: # cat /usr/src/linux/.config | grep NOTIFY CONFIG_FB_NOTIFY=y CONFIG_FSNOTIFY=y CONFIG_DNOTIFY=y CONFIG_INOTIFY_USER=y CONFIG_FANOTIFY=y This isn't a new experience (non-working nullmailer in a hardened VM); for at least the past year when I encountered this issue, I would 'sigh' and install Postfix instead. In a round-about way, this also implies the behaviour does not depend on GCC-4 transition to GCC-5, for one example. Thanks for looking!
Examining the file-structure more-closely: Working-server directory structure: # ls -al /var/spool/nullmailer/ ls: cannot access /var/spool/nullmailer/: No such file or directory non-Working-server directory structure: # ls -al /var/spool/nullmailer/ total 0 drwxr-xr-x 6 root root 69 Jun 22 09:30 . drwxr-xr-x 4 root root 46 Jun 20 12:00 .. drwx------ 2 root root 6 Jun 20 12:00 failed drwxrwx--- 2 nullmail nullmail 132 Jun 23 03:10 queue drwxr-xr-x 3 root root 26 Jun 22 09:30 service drwxrwx--- 2 nullmail nullmail 40 Jun 23 03:10 tmp prw-rw---- 1 nullmail nullmail 0 Jun 22 09:30 trigger SOLUTION: Re-configure svscan to supervise this different location: # ln -fs /var/spool/nullmailer/service /service/nullmailer # /etc/init.d/svscan restart Somewhere along the line, it seems /var/nullmailer got changed to /var/spool/nullmailer. Updating our internal documentation will address this problem. The only required change lies with svscan, and monitoring the correct directory. Thank-you,
