From 22d3e1840e271da6ad925fbe6d4560b6d196edf7 Mon Sep 17 00:00:00 2001
From: Guido Trotter <ultrotter@google.com>
Date: Fri, 28 Aug 2009 15:03:39 +0300
Subject: [PATCH] Confd: don't fail if the config doesn't load
Rather than quitting we'll just continue to poll the config at a slow
rate, hoping that sooner or later we'll get it back. This allows also
working on non-MC nodes, and smoothly transitioning from MC to non-MC,
other than dealing with a few temporary breakages.
Signed-off-by: Guido Trotter <ultrotter@google.com>
Reviewed-by: Iustin Pop <iustin@google.com>
---
daemons/ganeti-confd | 53 ++++++++++++++++++++++++++++++++------------
1 file changed, 39 insertions(+), 14 deletions(-)
diff --git a/daemons/ganeti-confd b/daemons/ganeti-confd
index cf56aa8f2..558d82a26 100755
--- a/daemons/ganeti-confd
+++ b/daemons/ganeti-confd
@@ -235,7 +235,7 @@ class ConfdConfigurationReloader(object):
try:
self.inotify_handler.enable()
except errors.InotifyError:
- raise errors.ConfdFatalError(err)
+ self.polling = True
try:
reloaded = self.processor.reader.Reload()
@@ -244,14 +244,13 @@ class ConfdConfigurationReloader(object):
else:
logging.debug("Skipped double config reload")
except errors.ConfigurationError:
- # transform a ConfigurationError in a fatal error, that will cause confd
- # to quit.
- raise errors.ConfdFatalError(err)
+ self.DisableConfd()
+ self.inotify_handler.disable()
+ return
# Reset the timer. If we're polling it will go to the polling rate, if
# we're not it will delay it again to its base safe timeout.
- self._DisableTimer()
- self._EnableTimer()
+ self._ResetTimer()
def _DisableTimer(self):
if self.timer_handle is not None:
@@ -268,17 +267,27 @@ class ConfdConfigurationReloader(object):
self.timer_handle = self.mainloop.scheduler.enter(
timeout, 1, self.OnTimer, [])
+ def _ResetTimer(self):
+ self._DisableTimer()
+ self._EnableTimer()
+
def OnTimer(self):
"""Function called when the timer fires
"""
self.timer_handle = None
+ reloaded = False
+ was_disabled = False
try:
- reloaded = self.processor.reader.Reload()
+ if self.processor.reader is None:
+ was_disabled = True
+ self.EnableConfd()
+ reloaded = True
+ else:
+ reloaded = self.processor.reader.Reload()
except errors.ConfigurationError:
- # transform a ConfigurationError in a fatal error, that will cause confd
- # to quit.
- raise errors.ConfdFatalError(err)
+ self.DisableConfd()
+ return
if self.polling and reloaded:
logging.info("Reloaded ganeti config")
@@ -286,19 +295,37 @@ class ConfdConfigurationReloader(object):
# We have reloaded the config files, but received no inotify event. If
# an event is pending though, we just happen to have timed out before
# receiving it, so this is not a problem, and we shouldn't alert
- if not self.notifier.check_events():
+ if not self.notifier.check_events() and not was_disabled:
logging.warning("Config file reload at timeout (inotify failure)")
elif self.polling:
# We're polling, but we haven't reloaded the config:
# Going back to inotify mode
logging.debug("Moving from polling mode to inotify mode")
self.polling = False
- self.inotify_handler.enable()
+ try:
+ self.inotify_handler.enable()
+ except errors.InotifyError:
+ self.polling = True
else:
logging.debug("Performed configuration check")
self._EnableTimer()
+ def DisableConfd(self):
+ """Puts confd in non-serving mode
+
+ """
+ logging.warning("Confd is being disabled")
+ self.processor.Disable()
+ self.polling = False
+ self._ResetTimer()
+
+ def EnableConfd(self):
+ self.processor.Enable()
+ logging.warning("Confd is being enabled")
+ self.polling = True
+ self._ResetTimer()
+
def CheckConfd(options, args):
"""Initial checks whether to run exit with a failure.
@@ -310,8 +337,6 @@ def CheckConfd(options, args):
print >> sys.stderr, "Need HMAC key %s to run" % constants.HMAC_CLUSTER_KEY
sys.exit(constants.EXIT_FAILURE)
- ssconf.CheckMasterCandidate(options.debug)
-
def ExecConfd(options, args):
"""Main confd function, executed with PID file held
--
GitLab