Restart vtbufferd on exit

Notably, if you strace it, it will trigger an exit(1).  There was at least one documented segmentation fault as well.

Buffer content is lost in such an event, but service remains running.

confluent_server/confluent/consoleserver.py

@@ -597,14 +597,21 @@ def _start_tenant_sessions(cfm):
                           event=log.Events.stacktrace)
     cfm.watch_nodecollection(_nodechange)
 
+running = True
+
+def run_buffer_daemon():
+    global _bufferdaemon
+    while running:
+        _bufferdaemon = subprocess.Popen(
+            ['/opt/confluent/bin/vtbufferd', 'confluent-vtbuffer'], bufsize=0, stdin=subprocess.DEVNULL,
+            stdout=subprocess.DEVNULL)
+        _bufferdaemon.wait()
 
 def initialize():
     global _tracelog
     global _bufferdaemon
     _tracelog = log.Logger('trace')
-    _bufferdaemon = subprocess.Popen(
-        ['/opt/confluent/bin/vtbufferd', 'confluent-vtbuffer'], bufsize=0, stdin=subprocess.DEVNULL,
-        stdout=subprocess.DEVNULL)
+    eventlet.spawn(run_buffer_daemon)
 
 def start_console_sessions():
     configmodule.hook_new_configmanagers(_start_tenant_sessions)