From 4f2b15dc8e32dc665da18c030451bd16690aa0df Mon Sep 17 00:00:00 2001
From: Jarrod Johnson <jbjohnso@us.ibm.com>
Date: Fri, 4 Apr 2014 18:40:45 -0400
Subject: [PATCH] First pass at automatic reconfiguration enablement

The ipmi plugin, at least, is not yet quite right.  Need to
continue debugging having a console session open, then changing
the bmc to a bad address, then changing it back.  I fixed some of
the easier exceptions, but it is clearly still getting quite confused
to the point where 3 or 4 cycles guarantees the console can not easily heal.
---
 confluent/consoleserver.py         | 18 +++++++++++++++++
 plugins/hardwaremanagement/ipmi.py | 32 ++++++++++++++++++++++--------
 2 files changed, 42 insertions(+), 8 deletions(-)

diff --git a/confluent/consoleserver.py b/confluent/consoleserver.py
index b76d362a..fab16fab 100644
--- a/confluent/consoleserver.py
+++ b/confluent/consoleserver.py
@@ -18,6 +18,8 @@ import random
 
 _handled_consoles = {}
 
+_genwatchattribs = frozenset(('console.method',))
+
 
 class _ConsoleHandler(object):
     def __init__(self, node, configmanager):
@@ -39,10 +41,17 @@ class _ConsoleHandler(object):
             self.shiftin = '0'
         self.users = {}
         self.connectstate = 'connecting'
+        self._attribwatcher = None
+        self._console = None
         eventlet.spawn(self._connect)
 
+    def _attribschanged(self, **kwargs):
+        eventlet.spawn(self._connect)
 
     def _connect(self):
+        if self._console:
+            self._console.close()
+            self._console = None
         self.connectstate = 'connecting'
         self._send_rcpts({'connectstate': self.connectstate})
         if self.reconnect:
@@ -51,6 +60,15 @@ class _ConsoleHandler(object):
         self._console = plugin.handle_path(
             "/nodes/%s/_console/session" % self.node,
             "create", self.cfgmgr)
+        if self._attribwatcher:
+            self.cfgmgr.remove_watcher(self._attribwatcher)
+            self._attribwatcher = None
+        if hasattr(self._console, "configattributes"):
+            attribstowatch = self._console.configattributes | _genwatchattribs
+        else:
+            attribstowatch = _genwatchattribs
+        self.cfgmgr.watch_attributes((self.node,), attribstowatch,
+                                     self._attribschanged)
         try:
             self._console.connect(self.get_console_output)
         except exc.TargetEndpointUnreachable:
diff --git a/plugins/hardwaremanagement/ipmi.py b/plugins/hardwaremanagement/ipmi.py
index 440519ab..44cb10fa 100644
--- a/plugins/hardwaremanagement/ipmi.py
+++ b/plugins/hardwaremanagement/ipmi.py
@@ -57,16 +57,18 @@ def get_conn_params(node, configdata):
         'port': 623,
     }
 
+_configattributes = ('secret.hardwaremanagementuser',
+             'secret.hardwaremanagementpassphrase',
+             'secret.ipmikg', 'hardwaremanagement.manager')
 
 class IpmiConsole(conapi.Console):
+    configattributes = frozenset(_configattributes)
+
     def __init__(self, node, config):
         crypt = config.decrypt
         config.decrypt = True
         self.broken = False
-        configdata = config.get_node_attributes([node],
-            ['secret.hardwaremanagementuser',
-             'secret.hardwaremanagementpassphrase',
-             'secret.ipmikg', 'hardwaremanagement.manager'])
+        configdata = config.get_node_attributes([node], _configattributes)
         connparams = get_conn_params(node, configdata[node])
         config.decrypt = crypt
         self.username = connparams['username']
@@ -77,10 +79,14 @@ class IpmiConsole(conapi.Console):
         self.connected = False
         # Cannot actually create console until 'connect', when we get callback
 
+    def __del__(self):
+        self.solconnection = None
+
     def handle_data(self, data):
         if type(data) == dict:
             disconnect = frozenset(('Session Disconnected', 'timeout'))
             if 'error' in data and data['error'] in disconnect:
+                self.solconnection = None
                 self.broken = True
                 self.error = data['error']
                 if self.connected:
@@ -92,6 +98,9 @@ class IpmiConsole(conapi.Console):
 
     def connect(self,callback):
         self.datacallback = callback
+        # we provide a weak reference to pyghmi as otherwise we'd
+        # have a circular reference and reference counting would never get
+        # out...
         try:
             self.solconnection = console.Console(bmc=self.bmc, port=self.port,
                                                 userid=self.username,
@@ -102,12 +111,22 @@ class IpmiConsole(conapi.Console):
                 w = eventlet.event.Event()
                 _ipmiwaiters.append(w)
                 w.wait()
+                if self.broken:
+                    break
             if self.broken:
                 raise exc.TargetEndpointUnreachable(self.error)
         except socket.gaierror as err:
             raise exc.TargetEndpointUnreachable(str(err))
 
 
+    def close(self):
+        if hasattr(self, 'solconnection') and self.solconnection is not None:
+            # break the circular reference here
+            self.solconnection.out_handler = None
+        self.solconnection = None
+        self.broken = True
+        self.error = "closed"
+
     def write(self, data):
         self.solconnection.send_data(data)
 
@@ -116,10 +135,7 @@ class IpmiIterator(object):
     def __init__(self, operator, nodes, element, cfg, inputdata):
         crypt = cfg.decrypt
         cfg.decrypt = True
-        configdata = cfg.get_node_attributes(nodes,
-            ['secret.hardwaremanagementuser',
-             'secret.hardwaremanagementpassphrase',
-             'secret.ipmikg', 'hardwaremanagement.manager'])
+        configdata = cfg.get_node_attributes(nodes, _configattributes)
         cfg.decrypt = crypt
         self.gpile = greenpool.GreenPile()
         for node in nodes: