3 # RunlevelAgent - acts as a heartbeat back to myplc reporting that the node is
4 # online and whether it is in boot or pre-boot run-level.
5 # This is useful to identify nodes that are behind a firewall, as well as to
6 # have the machine report run-time status both in safeboot and boot modes,
7 # so that it is immediately visible at myplc (gui or api).
18 CONFIG_FILE="/tmp/source/configuration"
19 SESSION_FILE="/etc/planetlab/session"
20 RLA_PID_FILE="/var/run/rla.pid"
22 def read_config_file(filename):
23 ## NOTE: text copied from BootManager.py
24 # TODO: unify this code to make it common. i.e. use ConfigParser module
26 vars_file= file(filename,'r')
28 for line in vars_file:
29 # if its a comment or a whitespace line, ignore
30 if line[:1] == "#" or string.strip(line) == "":
33 parts= string.split(line,"=")
35 print "Invalid line in vars file: %s" % line
39 name= string.strip(parts[0])
40 value= string.strip(parts[1])
45 print "Unable to read configuration vars."
50 sys.path = ['/etc/planetlab'] + sys.path
52 api_server_url = "https://" + plc_config.PLC_API_HOST + plc_config.PLC_API_PATH
55 vars = read_config_file(filename)
56 api_server_url = vars['BOOT_API_SERVER']
60 def __init__(self, username=None, password=None, **kwargs):
61 if 'session' in kwargs:
62 self.auth= { 'AuthMethod' : 'session',
63 'session' : kwargs['session'] }
65 if username==None and password==None:
66 self.auth = {'AuthMethod': "anonymous"}
68 self.auth = {'Username' : username,
69 'AuthMethod' : 'password',
70 'AuthString' : password}
72 def __init__(self, auth, url):
75 self.api = xmlrpclib.Server(self.url, verbose=False, allow_none=True)
77 def __getattr__(self, name):
78 method = getattr(self.api, name)
80 raise AssertionError("method does not exist")
82 return lambda *params : method(self.auth.auth, *params)
85 return self.api.__repr__()
87 def extract_from(filename, pattern):
88 f = os.popen("grep -E %s %s" % (pattern, filename))
89 val = f.read().strip()
92 def check_running(commandname):
93 f = os.popen("ps ax | grep -E %s | grep -v grep" % (commandname))
94 val = f.read().strip()
102 f = open(RLA_PID_FILE, 'w')
103 f.write("%s\n" % pid)
106 print "Uuuhhh.... this should not occur."
113 # Keep trying to authenticate session, waiting for NM to re-write the
114 # session file, or DNS to succeed, until AuthCheck succeeds.
117 f=open(SESSION_FILE,'r')
118 session_str=f.read().strip()
119 api = PLC(Auth(session=session_str), api_server_url)
120 # NOTE: What should we do if this call fails?
121 # TODO: handle dns failure here.
125 print "Retry in 30 seconds: ", os.popen("uptime").read().strip()
126 traceback.print_exc()
131 if len(sys.argv) > 2:
134 traceback.print_exc()
138 # NOTE: here we are inferring the runlevel by environmental
139 # observations. We know how this process was started by the
140 # given command line argument. Then in bootmanager
141 # runlevel, the bm.log gives information about the current
144 # call plc for current boot state?
145 # how long have we been running?
146 if env == "bootmanager":
147 bs_val = extract_from('/tmp/bm.log', "'Current boot state:'")
148 if len(bs_val) > 0: bs_val = bs_val.split()[-1]
149 ex_val = extract_from('/tmp/bm.log', 'Exception')
150 fs_val = extract_from('/tmp/bm.log', 'mke2fs')
151 bm_val = check_running("BootManager.py")
153 if bs_val in ['diag', 'diagnose', 'safeboot', 'disabled', 'disable']:
154 api.ReportRunlevel({'run_level' : 'safeboot'})
156 elif len(ex_val) > len("Exception"):
157 api.ReportRunlevel({'run_level' : 'failboot'})
159 elif len(fs_val) > 0 and len(bm_val) > 0:
160 api.ReportRunlevel({'run_level' : 'reinstall'})
163 api.ReportRunlevel({'run_level' : 'failboot'})
165 elif env == "production":
166 api.ReportRunlevel({'run_level' : 'boot'})
168 api.ReportRunlevel({'run_level' : 'failboot'})
171 print "reporting error: ", os.popen("uptime").read().strip()
172 traceback.print_exc()
175 # TODO: change to a configurable value
180 os.stat(RLA_PID_FILE)
181 f = os.popen("ps ax | grep RunlevelAgent | grep -Ev 'grep|vim' | awk '{print $1}' | wc -l")
187 os.unlink(RLA_PID_FILE)
198 pid = open(RLA_PID_FILE, 'r').read().strip()
200 # Try three different ways to kill the process. Just to be sure.
201 os.kill(int(pid), signal.SIGKILL)
202 os.system("pkill RunlevelAgent.py")
203 os.system("ps ax | grep RunlevelAgent | grep -v grep | awk '{print $1}' | xargs kill -9 ")
205 if __name__ == "__main__":
206 if "start" in sys.argv and not agent_running():
209 if "stop" in sys.argv and agent_running():