3 # Copyright (c) 2003 Intel Corporation
6 # Copyright (c) 2004-2006 The Trustees of Princeton University
10 import sys, os, traceback
15 from Exceptions import *
16 import notify_messages
17 import BootServerRequest
20 # all output is written to this file
21 BM_NODE_LOG= "/tmp/bm.log"
22 VARS_FILE = "configuration"
24 # the new contents of PATH when the boot manager is running
25 BIN_PATH= ('/usr/local/bin',
32 def read_configuration_file(filename):
33 # read in and store all variables in VARS_FILE into each line
34 # is in the format name=val (any whitespace around the = is
35 # removed. everything after the = to the end of the line is
38 vars_file= file(filename,'r')
40 for line in vars_file:
41 # if its a comment or a whitespace line, ignore
42 if line[:1] == "#" or string.strip(line) == "":
45 parts= string.split(line,"=")
48 raise Exception( "Invalid line in vars file: %s" % line )
50 name= string.strip(parts[0])
51 value= string.strip(parts[1])
52 value= value.replace("'", "") # remove quotes
53 value= value.replace('"', "") # remove quotes
58 raise Exception( "Unable to read configuration vars." )
60 # find out which directory we are running it, and set a variable
61 # for that. future steps may need to get files out of the bootmanager
63 current_dir= os.getcwd()
64 vars['BM_SOURCE_DIR']= current_dir
68 ##############################
71 format="%H:%M:%S(%Z) "
73 def __init__( self, OutputFilePath= None ):
75 self.OutputFile= open( OutputFilePath, "w")
76 self.OutputFilePath= OutputFilePath
78 print( "bootmanager log : Unable to open output file %r, continuing"%OutputFilePath )
83 vars = read_configuration_file(VARS_FILE)
86 self.LogEntry( str(e) )
89 def LogEntry( self, str, inc_newline= 1, display_screen= 1 ):
90 now=time.strftime(log.format, time.localtime())
92 self.OutputFile.write( now+str )
94 sys.stdout.write( now+str )
98 sys.stdout.write( "\n" )
100 self.OutputFile.write( "\n" )
103 self.OutputFile.flush()
105 def write( self, str ):
107 make log behave like a writable file object (for traceback
110 self.LogEntry( str, 0, 1 )
112 # bm log uploading is available back again, as of nodeconfig-5.0-2
113 def Upload( self, extra_file=None ):
115 upload the contents of the log to the server
117 if self.OutputFile is not None:
118 self.OutputFile.flush()
120 self.LogEntry( "Uploading logs to %s" % self.VARS['UPLOAD_LOG_SCRIPT'] )
122 self.OutputFile.close()
123 self.OutputFile= None
125 hostname= self.VARS['INTERFACE_SETTINGS']['hostname'] + "." + \
126 self.VARS['INTERFACE_SETTINGS']['domainname']
127 bs_request = BootServerRequest.BootServerRequest(self.VARS)
129 # this was working until f10
130 bs_request.MakeRequest(PartialPath = self.VARS['UPLOAD_LOG_SCRIPT'],
131 GetVars = None, PostVars = None,
132 DoSSL = True, DoCertCheck = True,
133 FormData = ["log=@" + self.OutputFilePath,
134 "hostname=" + hostname,
139 bs_request.MakeRequest(PartialPath = self.VARS['UPLOAD_LOG_SCRIPT'],
140 GetVars = None, PostVars = None,
141 DoSSL = True, DoCertCheck = True,
142 FormData = [('log',(pycurl.FORM_FILE, self.OutputFilePath)),
143 ("hostname",hostname),
145 if extra_file is not None:
146 # NOTE: for code-reuse, evoke the bash function 'upload_logs';
147 # by adding --login, bash reads .bash_profile before execution.
148 # Also, never fail, since this is an optional feature.
149 utils.sysexec_noerr( """bash --login -c "upload_logs %s" """ % extra_file, self)
152 ##############################
155 # file containing initial variables/constants
157 # the set of valid node run states
158 NodeRunStates = {'reinstall':None,
164 def __init__(self, log, forceState):
165 # override machine's current state from the command line
166 self.forceState = forceState
168 # the main logging point
171 # set to 1 if we can run after initialization
175 # this contains a set of information used and updated by each step
180 # not sure what the current PATH is set to, replace it with what
181 # we know will work with all the boot cds
182 os.environ['PATH']= string.join(BIN_PATH,":")
188 core boot manager logic.
190 the way errors are handled is as such: if any particular step
191 cannot continue or unexpectibly fails, an exception is thrown.
192 in this case, the boot manager cannot continue running.
194 these step functions can also return a 0/1 depending on whether
195 or not it succeeded. In the case of steps like ConfirmInstallWithUser,
196 a 0 is returned and no exception is thrown if the user chose not
197 to confirm the install. The same goes with the CheckHardwareRequirements.
198 If requriements not met, but tests were succesfull, return 0.
200 for steps that run within the installer, they are expected to either
201 complete succesfully and return 1, or throw an execption.
203 For exact return values and expected operations, see the comments
204 at the top of each of the invididual step functions.
207 def _nodeNotInstalled(message='MSG_NODE_NOT_INSTALLED'):
208 # called by the _xxxState() functions below upon failure
209 self.VARS['RUN_LEVEL']= 'failboot'
210 notify = getattr(notify_messages, message)
211 self.VARS['STATE_CHANGE_NOTIFY']= 1
212 self.VARS['STATE_CHANGE_NOTIFY_MESSAGE']= notify
213 raise BootManagerException, notify
216 # implements the boot logic, which consists of first
217 # double checking that the node was properly installed,
218 # checking whether someone added or changed disks, and
219 # then finally chain boots.
221 # starting the fallback/debug ssh daemon for safety:
222 # if the node install somehow hangs, or if it simply takes ages,
223 # we can still enter and investigate
225 StartDebug.Run(self.VARS, self.LOG, last_resort = False)
229 InstallInit.Run( self.VARS, self.LOG )
230 ret = ValidateNodeInstall.Run( self.VARS, self.LOG )
232 WriteModprobeConfig.Run( self.VARS, self.LOG )
233 WriteNetworkConfig.Run( self.VARS, self.LOG )
234 CheckForNewDisks.Run( self.VARS, self.LOG )
235 SendHardwareConfigToPLC.Run( self.VARS, self.LOG )
236 ChainBootNode.Run( self.VARS, self.LOG )
238 _nodeNotInstalled('MSG_NODE_FILESYSTEM_CORRUPT')
240 _nodeNotInstalled('MSG_NODE_MOUNT_FAILED')
242 _nodeNotInstalled('MSG_NODE_MISSING_KERNEL')
248 # starting the fallback/debug ssh daemon for safety:
249 # if the node install somehow hangs, or if it simply takes ages,
250 # we can still enter and investigate
252 StartDebug.Run(self.VARS, self.LOG, last_resort = False)
256 # implements the reinstall logic, which will check whether
257 # the min. hardware requirements are met, install the
258 # software, and upon correct installation will switch too
259 # 'boot' state and chainboot into the production system
260 if not CheckHardwareRequirements.Run( self.VARS, self.LOG ):
261 self.VARS['RUN_LEVEL']= 'failboot'
262 raise BootManagerException, "Hardware requirements not met."
265 InstallInit.Run( self.VARS, self.LOG )
266 InstallPartitionDisks.Run( self.VARS, self.LOG )
267 InstallBootstrapFS.Run( self.VARS, self.LOG )
268 InstallWriteConfig.Run( self.VARS, self.LOG )
269 InstallUninitHardware.Run( self.VARS, self.LOG )
270 self.VARS['BOOT_STATE']= 'boot'
271 self.VARS['STATE_CHANGE_NOTIFY']= 1
272 self.VARS['STATE_CHANGE_NOTIFY_MESSAGE']= \
273 notify_messages.MSG_INSTALL_FINISHED
274 UpdateBootStateWithPLC.Run( self.VARS, self.LOG )
278 # implements the new install logic, which will first check
279 # with the user whether it is ok to install on this
280 # machine, switch to 'reinstall' state and then invoke the reinstall
281 # logic. See reinstallState logic comments for further
283 if not ConfirmInstallWithUser.Run( self.VARS, self.LOG ):
285 self.VARS['BOOT_STATE']= 'reinstall'
288 def _debugRun(state='failboot'):
289 # implements debug logic, which starts the sshd and just waits around
290 self.VARS['RUN_LEVEL']=state
291 StartDebug.Run( self.VARS, self.LOG )
292 # fsck/mount fs if present, and ignore return value if it's not.
293 ValidateNodeInstall.Run( self.VARS, self.LOG )
296 # should never happen; log event
297 self.LOG.write( "\nInvalid BOOT_STATE = %s\n" % self.VARS['BOOT_STATE'])
300 # setup state -> function hash table
301 BootManager.NodeRunStates['reinstall'] = _reinstallRun
302 BootManager.NodeRunStates['boot'] = _bootRun
303 BootManager.NodeRunStates['safeboot'] = lambda : _debugRun('safeboot')
304 BootManager.NodeRunStates['disabled'] = lambda : _debugRun('disabled')
308 InitializeBootManager.Run( self.VARS, self.LOG )
309 ReadNodeConfiguration.Run( self.VARS, self.LOG )
310 AuthenticateWithPLC.Run( self.VARS, self.LOG )
311 UpdateLastBootOnce.Run( self.VARS, self.LOG )
312 StartRunlevelAgent.Run( self.VARS, self.LOG )
313 GetAndUpdateNodeDetails.Run( self.VARS, self.LOG )
315 # override machine's current state from the command line
316 if self.forceState is not None:
317 self.VARS['BOOT_STATE']= self.forceState
318 UpdateBootStateWithPLC.Run( self.VARS, self.LOG )
320 stateRun = BootManager.NodeRunStates.get(self.VARS['BOOT_STATE'],_badstateRun)
325 self.LOG.write( "\n\nKeyError while running: %s\n" % str(e) )
326 except BootManagerException, e:
327 self.LOG.write( "\n\nException while running: %s\n" % str(e) )
328 except BootManagerAuthenticationException, e:
329 self.LOG.write( "\n\nFailed to Authenticate Node: %s\n" % str(e) )
330 # sets /tmp/CANCEL_BOOT flag
331 StartDebug.Run(self.VARS, self.LOG )
332 # Return immediately b/c any other calls to API will fail
335 self.LOG.write( "\n\nImplementation Error\n")
336 traceback.print_exc(file=self.LOG.OutputFile)
337 traceback.print_exc()
342 except BootManagerException, e:
343 self.LOG.write( "\n\nException while running: %s\n" % str(e) )
345 self.LOG.write( "\n\nImplementation Error\n")
346 traceback.print_exc(file=self.LOG.OutputFile)
347 traceback.print_exc()
355 utils.prompt_for_breakpoint_mode()
357 # utils.breakpoint ("Entering BootManager::main")
359 # set to 1 if error occurred
362 # all output goes through this class so we can save it and post
363 # the data back to PlanetLab central
364 LOG= log( BM_NODE_LOG )
366 # NOTE: assume CWD is BM's source directory, but never fail
367 utils.sysexec_noerr("./setup_bash_history_scripts.sh", LOG)
369 LOG.LogEntry( "BootManager started at: %s" % \
370 time.strftime("%a, %d %b %Y %H:%M:%S +0000", time.gmtime()) )
376 if BootManager.NodeRunStates.has_key(fState):
379 LOG.LogEntry("FATAL: cannot force node run state to=%s" % fState)
382 traceback.print_exc(file=LOG.OutputFile)
383 traceback.print_exc()
386 LOG.LogEntry( "BootManager finished at: %s" % \
387 time.strftime("%a, %d %b %Y %H:%M:%S +0000", time.gmtime()) )
392 bm= BootManager(LOG,forceState)
394 LOG.LogEntry( "Unable to initialize BootManager." )
396 LOG.LogEntry( "Running version %s of BootManager." % bm.VARS['VERSION'] )
399 LOG.LogEntry( "\nDone!" );
401 LOG.LogEntry( "\nError occurred!" );
404 traceback.print_exc(file=LOG.OutputFile)
405 traceback.print_exc()
407 LOG.LogEntry( "BootManager finished at: %s" % \
408 time.strftime("%a, %d %b %Y %H:%M:%S +0000", time.gmtime()) )
414 if __name__ == "__main__":
415 error = main(sys.argv)