Kernel log collection stuck after node reboot

[tpdownes]

Our infrastructure management procedure sometimes involves a node reboot. We have observed that our kernel collection configuration:

      kernel-target = [
        { database = 'Logs', table = 'Kernel', priority = 'info' }
      ]

sometimes (but not always) stops progressing after a node reboot. A mitigation that works is the following steps:

1. Remove the file matching /mnt/collector/kernel_*
2. Restart the collector daemon

We are also independently looking at the solution mentioned in #653 (review).

[tpdownes]

I made this change which I have found to be a satisfactory reimplementation as suggsted by #653 (review). I note a couple things:

1. The kernel-target method includes trailing newlines in the messages (IMO this is a bug) while the journal-target does not (proper behavior)
2. The priority is no longer reported. I do not want to filter the journal target by priority since you have to include every priority match in the journal-fields setting (PRIORITY=6 PRIORITY=5 ...).

       journal-target = [
         # matches are optional and are parsed like MATCHES in journalctl.
         # If different fields are matched, only entries matching all terms are included.
         # If the same fields are matched, entries matching any term are included.
         # + can be added between to include a disjunction of terms.
         # See examples under man 1 journalctl
-        { matches = [ '_SYSTEMD_UNIT=kubelet.service' ], database = 'Logs', table = 'Kubelet' }
+        { matches = [ '_SYSTEMD_UNIT=kubelet.service' ], database = 'Logs', table = 'Kubelet' },
+        { matches = [ '_TRANSPORT=kernel' ], database = 'Logs', table = 'Kernel' }
       ]

-      kernel-target = [
-        { database = 'Logs', table = 'Kernel', priority = 'info' }
-      ]