watcher: try to restart the master if down

Bugs in either our code or in associated libraries can bring the master daemon
down, and this (due to the 2.0 architecture) stops all work on the cluster.

Since the watcher already does periodic checks on the cluster, we modify
it to try to start the master automatically in case of failures to
connect. This will be tried only once per cycle.

Also, in this case, we modify the code so that the watcher status file
is not updated - its timestamp will reflect thus the time of last
successful connection to the master.

Side note: the except errors.ConfigurationError part could be cleaned
up, since in 2.0 we don't usually get that directly, and if we do it's
an error and we shouldn't touch the file anyway; but that is not a rc5
change.

Signed-off-by: Iustin Pop <iustin@google.com>

diff --git a/daemons/ganeti-watcher b/daemons/ganeti-watcher
index 8745f2d..42a2eaf 100755 (executable)
--- a/daemons/ganeti-watcher
+++ b/daemons/ganeti-watcher
@@ -39,6 +39,7 @@ from ganeti import serializer
 from ganeti import errors
 from ganeti import opcodes
 from ganeti import cli
+from ganeti import luxi
 
 
 MAXTRIES = 5
@@ -69,6 +70,16 @@ def Indent(s, prefix='| '):
   return "%s%s\n" % (prefix, ('\n' + prefix).join(s.splitlines()))
 
 
+def StartMaster():
+  """Try to start the master daemon.
+
+  """
+  result = utils.RunCmd(['ganeti-masterd'])
+  if result.failed:
+    logging.error("Can't start the master daemon: output '%s'", result.output)
+  return not result.failed
+
+
 class WatcherState(object):
   """Interface to a state file recording restart attempts.
 
@@ -441,6 +452,7 @@ def main():
   utils.SetupLogging(constants.LOG_WATCHER, debug=options.debug,
                      stderr_logging=options.debug)
 
+  update_file = True
   try:
     notepad = WatcherState()
     try:
@@ -448,7 +460,17 @@ def main():
         client = cli.GetClient()
       except errors.OpPrereqError:
         # this is, from cli.GetClient, a not-master case
+        logging.debug("Not on master, exiting")
         sys.exit(constants.EXIT_SUCCESS)
+      except luxi.NoMasterError, err:
+        logging.warning("Master seems to be down (%s), trying to restart",
+                        str(err))
+        if not StartMaster():
+          logging.critical("Can't start the master, exiting")
+          update_file = False
+          sys.exit(constants.EXIT_FAILURE)
+        # else retry the connection
+        client = cli.GetClient()
 
       try:
         watcher = Watcher(options, notepad)
@@ -458,7 +480,10 @@ def main():
 
       watcher.Run()
     finally:
-      notepad.Save()
+      if update_file:
+        notepad.Save()
+      else:
+        logging.debug("Not updating status file due to failure")
   except SystemExit:
     raise
   except NotMasterError: