3 # This script is used to manipulate the operational state of nodes in
4 # different node groups. These are basically set operations on nodes via the
7 # Take the ng name as an argument....
9 # * get a list of nodes in the given nodegroup.
10 # * set some or all in the set to rins.
12 # * do something else to them all.
16 api = plc.getAuthAPI()
20 from config import config as cfg
21 import config as configmodule
22 from optparse import OptionParser
24 from nodecommon import *
25 from nodequery import verify,query_to_dict,node_select
27 from unified_model import *
33 import bootman # debug nodes
34 import monitor # down nodes with pcu
35 import reboot # down nodes without pcu
36 from emailTxt import mailtxt
41 def __init__(self, fbnode):
44 def _send_pcunotice(self, host):
46 args['hostname'] = host
48 args['pcu_id'] = plc.getpcu(host)['pcu_id']
52 m = PersistMessage(host, mailtxt.pcudown_one[0] % args,
53 mailtxt.pcudown_one[1] % args, True, db='pcu_persistmessages')
55 loginbase = plc.siteId(host)
56 m.send([policy.TECHEMAIL % loginbase])
59 # TODO: It should be possible to diagnose the various conditions of
60 # the PCU here, and send different messages as appropriate.
61 if self.fbnode['pcu'] == "PCU":
62 self.action = "reboot.reboot('%s')" % host
64 pflags = PersistFlags(host, 2*60*60*24, db='pcu_persistflags')
65 if not pflags.getRecentFlag('pcutried'):
66 pflags.setRecentFlag('pcutried')
68 ret = reboot.reboot(host)
74 print traceback.print_exc(); print e
76 # NOTE: this failure could be an implementation issue on
77 # our end. So, extra notices are confusing...
78 # self._send_pcunotice(host)
80 pflags.setRecentFlag('pcufailed')
84 # we've tried the pcu recently, but it didn't work,
85 # so did we send a message about it recently?
86 if not pflags.getRecentFlag('pcumessagesent'):
88 self._send_pcunotice(host)
90 pflags.setRecentFlag('pcumessagesent')
92 # NOTE: this will result in just one message sent at a time.
101 def mail(self, host):
103 # Reset every 4 weeks or so
104 pflags = PersistFlags(host, 27*60*60*24, db='mail_persistflags')
105 if not pflags.getRecentFlag('endrecord'):
106 node_end_record(host)
107 pflags.setRecentFlag('endrecord')
110 # Then in either case, run monitor.reboot()
111 self.action = "monitor.reboot('%s')" % host
113 return monitor.reboot(host)
115 print traceback.print_exc(); print e
118 class RebootDebug(Reboot):
120 def direct(self, host):
121 self.action = "bootman.reboot('%s', config, None)" % host
122 return bootman.reboot(host, config, None)
124 class RebootBoot(Reboot):
126 def direct(self, host):
127 self.action = "bootman.reboot('%s', config, 'reboot')" % host
128 return bootman.reboot(host, config, 'reboot')
130 class RebootDown(Reboot):
132 def direct(self, host):
134 return False # this always fails, since the node will be down.
136 def set_node_to_rins(host, fb):
138 node = api.GetNodes(host, ['boot_state', 'last_contact', 'last_updated', 'date_created'])
139 record = {'observation' : node[0],
140 'model' : 'USER_REQUEST',
141 'action' : 'api.UpdateNode(%s, {"boot_state" : "rins"})' % host,
142 'time' : time.time()}
143 l = Log(host, record)
145 ret = api.UpdateNode(host, {'boot_state' : 'rins'})
147 # it's nice to see the current status rather than the previous status on the console
148 node = api.GetNodes(host)[0]
150 print "%-2d" % (i-1), nodegroup_display(node, fb)
153 print "FAILED TO UPDATE NODE BOOT STATE : %s" % host
158 rebootlog = database.dbLoad("rebootlog")
160 rebootlog = LogRoll()
162 parser = OptionParser()
163 parser.set_defaults(nodegroup=None,
179 parser.add_option("", "--node", dest="node", metavar="nodename.edu",
180 help="A single node name to add to the nodegroup")
181 parser.add_option("", "--nodelist", dest="nodelist", metavar="list.txt",
182 help="Use all nodes in the given file for operation.")
183 parser.add_option("", "--nodegroup", dest="nodegroup", metavar="NodegroupName",
184 help="Specify a nodegroup to perform actions on")
185 parser.add_option("", "--nodeselect", dest="nodeselect", metavar="querystring",
186 help="Specify a query to perform on findbad db")
188 parser.add_option("", "--verbose", dest="verbose", action="store_true",
189 help="Extra debug output messages.")
190 parser.add_option("", "--nosetup", dest="nosetup", action="store_true",
191 help="Do not perform the orginary setup phase.")
193 parser.add_option("", "--skip", dest="skip",
194 help="Number of machines to skip on the input queue.")
195 parser.add_option("", "--timewait", dest="timewait",
196 help="Minutes to wait between iterations of 10 nodes.")
198 parser.add_option("", "--stopselect", dest="stopselect", metavar="",
199 help="The select string that must evaluate to true for the node to be considered 'done'")
201 parser.add_option("", "--stopkey", dest="stopkey", metavar="",
203 parser.add_option("", "--stopvalue", dest="stopvalue", metavar="",
206 parser.add_option("", "--findbad", dest="findbad", action="store_true",
207 help="Re-run findbad on the nodes we're going to check before acting.")
208 parser.add_option("", "--force", dest="force", action="store_true",
209 help="Force action regardless of previous actions/logs.")
210 parser.add_option("", "--rins", dest="rins", action="store_true",
211 help="Set the boot_state to 'rins' for all nodes.")
212 parser.add_option("", "--reboot", dest="reboot", action="store_true",
213 help="Actively try to reboot the nodes, keeping a log of actions.")
214 #config = config(parser)
218 # COLLECT nodegroups, nodes and node lists
220 ng = api.GetNodeGroups({'name' : config.nodegroup})
221 nodelist = api.GetNodes(ng[0]['node_ids'])
222 hostnames = [ n['hostname'] for n in nodelist ]
224 if config.node or config.nodelist:
225 if config.node: hostnames = [ config.node ]
226 else: hostnames = config.getListFromFile(config.nodelist)
228 if config.nodeselect:
229 hostnames = node_select(config.nodeselect)
232 # rerun findbad with the nodes in the given nodes.
234 configmodule.setFileFromList(file, hostnames)
235 os.system("./findbad.py --cachenodes --debug=0 --dbname=findbad --increment --nodelist %s" % file)
237 fb = database.dbLoad("findbad")
241 for host in hostnames:
243 #if 'echo' in host or 'hptest-1' in host: continue
246 node = api.GetNodes(host)[0]
248 print traceback.print_exc();
249 print "FAILED GETNODES for host: %s" % host
252 print "%-2d" % i, nodegroup_display(node, fb)
254 if i < int(config.skip): continue
256 if config.stopselect:
257 dict_query = query_to_dict(config.stopselect)
258 fbnode = fb['nodes'][host]['values']
259 observed_state = get_current_state(fbnode)
261 if verify(dict_query, fbnode) and observed_state != "dbg ":
262 # evaluates to true, therefore skip.
263 print "%s evaluates true for %s ; skipping..." % ( config.stopselect, host )
266 if config.stopkey and config.stopvalue:
267 fbnode = fb['nodes'][host]['values']
268 observed_state = get_current_state(fbnode)
270 if config.stopkey in fbnode:
271 if config.stopvalue in fbnode[config.stopkey] and observed_state != "dbg ":
272 print "%s has stopvalue; skipping..." % host
275 print "stopkey %s not in fbnode record for %s; skipping..." % (config.stopkey, host)
279 if not config.force and rebootlog.find(host, {'action' : ".*reboot"}, 60*60*2):
280 print "recently rebooted %s. skipping... " % host
285 fbnode = fb['nodes'][host]['values']
286 observed_state = get_current_state(fbnode)
288 if observed_state == "dbg ":
289 o = RebootDebug(fbnode)
291 elif observed_state == "boot" :
293 l = set_node_to_rins(host, fb)
294 if l: rebootlog.add(l)
296 o = RebootBoot(fbnode)
298 elif observed_state == "down":
300 l = set_node_to_rins(host, fb)
301 if l: rebootlog.add(l)
303 o = RebootDown(fbnode)
307 record = {'observation' : "DIRECT_SUCCESS: %s" % observed_state,
310 'time' : time.time()}
312 record = {'observation' : "PCU_SUCCESS: %s" % observed_state,
315 'time' : time.time()}
317 record = {'observation' : "MAIL_SUCCESS: %s" % observed_state,
320 'time' : time.time()}
322 record = {'observation' : "REBOOT_FAILED: %s" % observed_state,
323 'action' : "log failure",
325 'time' : time.time()}
327 print "ALL METHODS OF RESTARTING %s FAILED" % host
329 args['hostname'] = host
330 m = PersistMessage(host, "ALL METHODS FAILED for %(hostname)s" % args,
331 "CANNOT CONTACT", False, db='suspect_persistmessages')
333 m.send(['monitor-list@lists.planet-lab.org'])
335 l = Log(host, record)
338 except KeyboardInterrupt:
339 print "Killed by interrupt"
342 print traceback.print_exc();
343 print "Continuing..."
347 print "Saving rebootlog"
348 database.dbDump("rebootlog", rebootlog)
349 wait_time = int(config.timewait)
350 print "Sleeping %d minutes" % wait_time
352 print "Minutes slept: ",
354 while ti < wait_time:
362 print "Saving rebootlog"
363 database.dbDump("rebootlog", rebootlog)