def load_config(opts, path, env_var):
'''
Attempts to update ``opts`` dict by parsing either the file described by
``path`` or the environment variable described by ``env_var`` as YAML.
'''
if not path or not os.path.isfile(path):
path = os.environ.get(env_var, path)
# If the configuration file is missing, attempt to copy the template,
# after removing the first header line.
if not os.path.isfile(path):
template = '{0}.template'.format(path)
if os.path.isfile(template):
with open(path, 'w') as out:
with open(template, 'r') as f:
f.readline() # skip first line
out.write(f.read())
if os.path.isfile(path):
try:
opts.update(_read_conf_file(path))
opts['conf_file'] = path
except Exception as e:
import salt.log
msg = 'Error parsing configuration file: {0} - {1}'
if salt.log.is_console_configured():
log.warn(msg.format(path, e))
else:
print(msg.format(path, e))
else:
log.debug('Missing configuration file: {0}'.format(path))
def grain_match(self, tgt):
'''
Reads in the grains glob match
'''
log.debug('grains target: {0}'.format(tgt))
comps = tgt.rsplit(':', 1)
if len(comps) != 2:
log.error('Got insufficient arguments for grains match '
'statement from master')
return False
match = self._traverse_dict(self.opts['grains'], comps[0])
if match == {}:
log.error('Targeted grain "{0}" not found'.format(comps[0]))
return False
if isinstance(match, dict):
log.error('Targeted grain "{0}" must correspond to a list, '
'string, or numeric value'.format(comps[0]))
return False
if isinstance(match, list):
# We are matching a single component to a single list member
for member in match:
if fnmatch.fnmatch(str(member).lower(), comps[1].lower()):
return True
return False
return bool(fnmatch.fnmatch(str(match).lower(), comps[1].lower()))
def start(self):
log.debug("starting monitor with {} task{}".format(len(self.tasks), "" if len(self.tasks) == 1 else "s"))
if self.tasks:
for task in self.tasks:
threading.Thread(target=task.run).start()
else:
log.error("no monitor tasks to run")
def _handle_aes(self, load):
'''
Takes the aes encrypted load, decrypts is and runs the encapsulated
instructions
'''
try:
data = self.crypticle.loads(load)
except AuthenticationError:
self.authenticate()
data = self.crypticle.loads(load)
# Verify that the publication is valid
if 'tgt' not in data or 'jid' not in data or 'fun' not in data \
or 'arg' not in data:
return
# Verify that the publication applies to this minion
if 'tgt_type' in data:
if not getattr(self.matcher,
'{0}_match'.format(data['tgt_type']))(data['tgt']):
return
else:
if not self.matcher.glob_match(data['tgt']):
return
# If the minion does not have the function, don't execute,
# this prevents minions that could not load a minion module
# from returning a predictable exception
#if data['fun'] not in self.functions:
# return
if 'user' in data:
log.info(('User {0[user]} Executing command {0[fun]} with jid '
'{0[jid]}'.format(data)))
else:
log.info(('Executing command {0[fun]} with jid {0[jid]}'
.format(data)))
log.debug('Command details {0}'.format(data))
self._handle_decoded_payload(data)
def stop(self,
signal,
frame):
'''
we override stop() to brake our main loop
and have a pretty log message
'''
log.info("received signal {0}".format(signal))
# if we have running workers, run through all and join() the ones
# that have finished. if we still have running workers after that,
# wait 5 secs for the rest and then exit. Maybe we should improv
# this a litte bit more
if( len(self.running_workers) > 0 ):
clean_workers = []
for count in range(0, 2):
for worker in self.running_workers:
if worker.isAlive():
clean_workers.append(worker)
else:
worker.join()
log.debug("joined worker #{0}".format(worker.getName()))
if( len(clean_workers) > 0 ):
log.info("waiting 5secs for remaining workers..")
time.sleep(5)
else:
break
log.info("salt-eventsd has shut down")
# leave the cleanup to the supers stop
super(SaltEventsDaemon, self).stop(signal, frame)
def _dism(action,
image=None):
'''
Run a DISM servicing command on the given image.
'''
command='dism {0} {1}'.format(
'/Image:{0}'.format(image) if image else '/Online',
action
)
ret = {'action': action,
'image': image,
'result': True,
'message': ''}
output = __salt__['cmd.run'](command, ignore_retcode=True)
if not re.search('The operation completed successfully.', output):
ret['result'] = False
ret['message'] = re.search(
'(Error: \d+\r?\n\r?\n([^\r\n]+\r?\n)+\r?\nThe DISM log file can be found at [^\r\n]\r?\n)',
output,
re.MULTILINE
).group(1)
log.exception('DISM command \'{0}\' on image {1} failed: {2}'.format(action, image, ret['message']))
else:
ret['message'] = output
log.debug('DISM command \'{0}\' on image {1} completed with the following output: {2}'.format(action, image, output))
return ret
def __init__(self,
tgt='',
expr_form='glob',
env=None,
use_cached_grains=True,
use_cached_pillar=True,
grains_fallback=True,
pillar_fallback=True,
opts=None):
log.debug('New instance of {0} created.'.format(
self.__class__.__name__))
if opts is None:
log.error('{0}: Missing master opts init arg.'.format(
self.__class__.__name__))
raise SaltException('{0}: Missing master opts init arg.'.format(
self.__class__.__name__))
else:
self.opts = opts
self.tgt = tgt
self.expr_form = expr_form
self.env = env
self.use_cached_grains = use_cached_grains
self.use_cached_pillar = use_cached_pillar
self.grains_fallback = grains_fallback
self.pillar_fallback = pillar_fallback
log.debug('Init settings: tgt: \"{0}\", expr_form: \"{1}\", env: \"{2}\", use_cached_grains: {3}, use_cached_pillar: {4}, grains_fallback: {5}, pillar_fallback: {6}'.format(tgt, expr_form, env, use_cached_grains, use_cached_pillar, grains_fallback, pillar_fallback))
def _get_cached_minion_data(self, *minion_ids):
# Return two separate dicts of cached grains and pillar data of the
# minions
grains = dict([(minion_id, {}) for minion_id in minion_ids])
pillars = grains.copy()
if not self.opts.get('minion_data_cache', False):
log.debug('Skipping cached data because minion_data_cache is not '
'enabled.')
return grains, pillars
mdir = os.path.join(self.opts['cachedir'], 'minions')
try:
for minion_id in minion_ids:
if not salt.utils.verify.valid_id(self.opts, minion_id):
continue
path = os.path.join(mdir, minion_id, 'data.p')
if os.path.isfile(path):
with salt.utils.fopen(path, 'rb') as fp_:
mdata = self.serial.loads(fp_.read())
if mdata.get('grains', False):
grains[minion_id] = mdata['grains']
if mdata.get('pillar', False):
pillars[minion_id] = mdata['pillar']
except (OSError, IOError):
return grains, pillars
return grains, pillars
def _get_cached_minion_data(self, *minion_ids):
# Return two separate dicts of cached grains and pillar data of the minions
grains = dict([(minion_id, {}) for minion_id in minion_ids])
pillars = grains.copy()
if not self.opts.get('minion_data_cache', False):
log.debug('Skipping cached data because minion_data_cache is not enabled.')
return grains, pillars
serial = salt.payload.Serial(self.opts)
mdir = os.path.join(self.opts['cachedir'], 'minions')
# salt.utils.verify.valid_id has changed in git development to require opts arg
valid_id_args = inspect.getargspec(salt.utils.verify.valid_id).args
log.debug('salt.utils.verify.valid_id accepts args: {0}'.format(valid_id_args))
try:
for minion_id in minion_ids:
if 'opts' in valid_id_args:
if not salt.utils.verify.valid_id(self.opts, minion_id):
continue
else:
if not salt.utils.verify.valid_id(self.opts, minion_id):
continue
path = os.path.join(mdir, minion_id, 'data.p')
if os.path.isfile(path):
with salt.utils.fopen(path) as fp_:
mdata = serial.loads(fp_.read())
if mdata.get('grains', False):
grains[minion_id] = mdata['grains']
if mdata.get('pillar', False):
pillars[minion_id] = mdata['pillar']
except (OSError, IOError):
return grains, pillars
return grains, pillars
def __init__(self, parser):
self.parser = parser
log.debug("TODO: load more manager here")
self.manager = LibvirtManager()
self.image = ImageService()
self.flavor = FlavorService()
def ext_pillar(minion_id, # pylint: disable=W0613
pillar, # pylint: disable=W0613
config_file):
'''
Execute LDAP searches and return the aggregated data
'''
if os.path.isfile(config_file):
try:
#open(config_file, 'r') as raw_config:
config = _render_template(config_file) or {}
opts = yaml.safe_load(config) or {}
opts['conf_file'] = config_file
except Exception as err:
import salt.log
msg = 'Error parsing configuration file: {0} - {1}'
if salt.log.is_console_configured():
log.warn(msg.format(config_file, err))
else:
print(msg.format(config_file, err))
else:
log.debug('Missing configuration file: {0}'.format(config_file))
data = {}
for source in opts['search_order']:
config = opts[source]
result = _do_search(config)
print('source {0} got result {1}'.format(source, result))
if result:
data = _result_to_dict(data, result, config)
return data
def __init__(self, opts, log_queue=None):
'''
starts the timer and inits the cache itself
'''
super(ConnectedCache, self).__init__(log_queue=log_queue)
log.debug('ConCache initializing...')
# the possible settings for the cache
self.opts = opts
# the actual cached minion ids
self.minions = []
self.cache_sock = os.path.join(self.opts['sock_dir'], 'con_cache.ipc')
self.update_sock = os.path.join(self.opts['sock_dir'], 'con_upd.ipc')
self.upd_t_sock = os.path.join(self.opts['sock_dir'], 'con_timer.ipc')
self.cleanup()
# the timer provides 1-second intervals to the loop in run()
# to make the cache system most responsive, we do not use a loop-
# delay which makes it hard to get 1-second intervals without a timer
self.timer_stop = Event()
self.timer = CacheTimer(self.opts, self.timer_stop)
self.timer.start()
self.running = True
def get_minion_grains(self):
'''
Get grains data for the targeted minions, either by fetching the
cached minion data on the master, or by fetching the grains
directly on the minion.
By default, this function tries hard to get the pillar data:
- Try to get the cached minion grains if the master
has minion_data_cache: True
- If the grains data for the minion is cached, use it.
- If there is no cached grains data for a minion,
then try to get the minion grains directly from the minion.
'''
minion_grains = {}
minion_ids = self._tgt_to_list()
if any(arg for arg in [self.use_cached_grains, self.grains_fallback]):
log.debug('Getting cached minion data.')
cached_minion_grains, cached_minion_pillars = self._get_cached_minion_data(*minion_ids)
else:
cached_minion_grains = {}
log.debug('Getting minion grain data for: {0}'.format(minion_ids))
minion_grains = self._get_minion_grains(
*minion_ids,
cached_grains=cached_minion_grains)
return minion_grains
def immutable_encoder(obj):
log.debug('IMMUTABLE OBJ: {0}'.format(obj))
if isinstance(obj, immutabletypes.ImmutableDict):
return dict(obj)
if isinstance(obj, immutabletypes.ImmutableList):
return list(obj)
if isinstance(obj, immutabletypes.ImmutableSet):
return set(obj)
def run(self):
'''
Gather currently connected minions and update the cache
'''
new_mins = list(salt.utils.minions.CkMinions(self.opts).connected_ids())
cc = cache_cli(self.opts)
cc.get_cached()
cc.put_cache([new_mins])
log.debug('ConCache CacheWorker update finished')
def get_cached_mine_data(self):
'''
Get cached mine data for the targeted minions.
'''
mine_data = {}
minion_ids = self._tgt_to_list()
log.debug('Getting cached mine data for: {0}'.format(minion_ids))
mine_data = self._get_cached_mine_data(*minion_ids)
return mine_data
def _linux_gpu_data():
'''
num_gpus: int
gpus:
- vendor: nvidia|amd|ati|...
model: string
'''
# dominant gpu vendors to search for (MUST be lowercase for matching below)
known_vendors = ['nvidia', 'amd', 'ati', 'intel']
devs = []
try:
lspci_out = __salt__['cmd.run']('lspci -vmm')
cur_dev = {}
error = False
for line in lspci_out.splitlines():
# check for record-separating empty lines
if line == '':
if cur_dev.get('Class', '') == 'VGA compatible controller':
devs.append(cur_dev)
# XXX; may also need to search for "3D controller"
cur_dev = {}
continue
if re.match(r'^\w+:\s+.*', line):
key, val = line.split(':', 1)
cur_dev[key.strip()] = val.strip()
else:
error = True
log.debug('Unexpected lspci output: \'{0}\''.format(line))
if error:
log.warn(
'Error loading grains, unexpected linux_gpu_data output, '
'check that you have a valid shell configured and '
'permissions to run lspci command'
)
except OSError:
pass
gpus = []
for gpu in devs:
vendor_strings = gpu['Vendor'].lower().split()
# default vendor to 'unknown', overwrite if we match a known one
vendor = 'unknown'
for name in known_vendors:
# search for an 'expected' vendor name in the list of strings
if name in vendor_strings:
vendor = name
break
gpus.append({'vendor': vendor, 'model': gpu['Device']})
grains = {}
grains['num_gpus'] = len(gpus)
grains['gpus'] = gpus
return grains
def _tgt_to_list(self):
# Return a list of minion ids that match the target and expr_form
minion_ids = []
ckminions = salt.utils.minions.CkMinions(self.opts)
minion_ids = ckminions.check_minions(self.tgt, self.expr_form)
if len(minion_ids) == 0:
log.debug('No minions matched for tgt="{0}" and expr_form="{1}"'.format(self.tgt, self.expr_form))
return {}
log.debug('Matching minions for tgt="{0}" and expr_form="{1}": {2}'.format(self.tgt, self.expr_form, minion_ids))
return minion_ids
def _get_live_minion_grains(self, minion_ids):
# Returns a dict of grains fetched directly from the minions
log.debug('Getting live grains for minions: "{0}"'.format(minion_ids))
client = salt.client.get_local_client(self.opts['conf_file'])
ret = client.cmd(
','.join(minion_ids),
'grains.items',
timeout=self.opts['timeout'],
expr_form='list')
return ret
def renew(self):
'''
compares the current minion list against the ips
connected on the master publisher port and updates
the minion list accordingly
'''
log.debug('ConCache renewing minion cache')
new_mins = list(salt.utils.minions.CkMinions(self.opts).connected_ids())
self.minions = new_mins
log.debug('ConCache received {0} minion ids'.format(len(new_mins)))
def _dmidecode_data(regex_dict):
'''
Parse the output of dmidecode in a generic fashion that can
be used for the multiple system types which have dmidecode.
'''
ret = {}
if 'proxyminion' in __opts__:
return {}
# No use running if dmidecode/smbios isn't in the path
if salt.utils.which('dmidecode'):
out = __salt__['cmd.run']('dmidecode')
elif salt.utils.which('smbios'):
out = __salt__['cmd.run']('smbios')
else:
log.debug(
'The `dmidecode` binary is not available on the system. GPU '
'grains will not be available.'
)
return ret
for section in regex_dict:
section_found = False
# Look at every line for the right section
for line in out.splitlines():
if not line:
continue
# We've found it, woohoo!
if re.match(section, line):
section_found = True
continue
if not section_found:
continue
# Now that a section has been found, find the data
for item in regex_dict[section]:
# Examples:
# Product Name: 64639SU
# Version: 7LETC1WW (2.21 )
regex = re.compile(r'\s+{0}\s+(.*)$'.format(item))
grain = regex_dict[section][item]
# Skip to the next iteration if this grain
# has been found in the dmidecode output.
if grain in ret:
continue
match = regex.match(line)
# Finally, add the matched data to the grains returned
if match:
ret[grain] = match.group(1).strip()
return ret
def __connect(self):
'''
Connect to the Jabber server.
'''
log.debug('connecting to {}:{}'.format(*self.server_addr))
if not self.connect(self.server_addr):
raise JabberError('connect failed: {}:{}'
.format(*self.server_addr))
# Process Jabber messages forever in a background thread
self.process(block=False)
def get_elb_lbs():
"""
Returns a dictionary of load balancer names as keys
each with their respective attributes
"""
# attributes to extract from the load balancer boto objects
# this could possibly be a named argument too
extract_attrs = ['scheme', 'dns_name', 'vpc_id', 'name', 'security_groups']
try:
instance_metadata = boto.utils.get_instance_metadata(timeout=5, num_retries=2)
except Exception as e:
log.exception("Error getting ELB names: {}".format(e))
return {'custom_grain_error': True}
# Setup the lbs grain
lbs_grain = {'lbs': {}}
# Collect details about this instance
vpc_id = instance_metadata['network']['interfaces']['macs'].values()[0]['vpc-id']
region = instance_metadata['placement']['availability-zone'][:-1]
# Collect load balancers of this instance (in the same vpc)
try:
elb_connection = boto.ec2.elb.connect_to_region(region)
# find load balancers by vpc_id
all_lbs = [lb for lb in elb_connection.get_all_load_balancers()
if lb.vpc_id == vpc_id]
log.debug('all lbs before filtering by instance id: {}'.format(all_lbs))
# further filter the load balancers by instance id
lbs = [lb for lb in all_lbs for inst in lb.instances
if inst.id == instance_metadata['instance-id']]
# initialise and populate the output of load balancers
out = {}
[out.update({l.name: {}}) for l in lbs]
[out[l.name].update({attr: getattr(l, attr, None)})
for attr in extract_attrs for l in lbs]
if not out:
# This loglevel could perhaps be adjusted to something more visible
log.warning("No ELBs found for this instance, this is unusual, "
"but we will not break highstate")
lbs_grain['lbs'] = out
except Exception as e:
# This prints a user-friendly error with stacktrace
log.exception("Error getting ELB names: {}".format(e))
return {'custom_grain_error': True}
return lbs_grain
def secure(self):
'''
secure the sockets for root-only access
'''
log.debug('ConCache securing sockets')
if os.path.exists(self.cache_sock):
os.chmod(self.cache_sock, 0o600)
if os.path.exists(self.update_sock):
os.chmod(self.update_sock, 0o600)
if os.path.exists(self.upd_t_sock):
os.chmod(self.upd_t_sock, 0o600)
def cleanup(self):
'''
remove sockets on shutdown
'''
log.debug('ConCache cleaning up')
if os.path.exists(self.cache_sock):
os.remove(self.cache_sock)
if os.path.exists(self.update_sock):
os.remove(self.update_sock)
if os.path.exists(self.upd_t_sock):
os.remove(self.upd_t_sock)
def get_id():
'''
Guess the id of the minion.
- If socket.getfqdn() returns us something other than localhost, use it
- Check /etc/hosts for something that isn't localhost that maps to 127.*
- Look for a routeable / public IP
- A private IP is better than a loopback IP
- localhost may be better than killing the minion
'''
log.debug('Guessing ID. The id can be explicitly in set {0}'
.format('/etc/salt/minion'))
fqdn = socket.getfqdn()
if 'localhost' != fqdn:
log.info('Found minion id from getfqdn(): {0}'.format(fqdn))
return fqdn, False
# Can /etc/hosts help us?
try:
# TODO Add Windows host file support
with open('/etc/hosts') as f:
line = f.readline()
while line:
names = line.split()
ip = names.pop(0)
if ip.startswith('127.'):
for name in names:
if name != 'localhost':
log.info('Found minion id in hosts file: {0}'
.format(name))
return name, False
line = f.readline()
except Exception:
pass
# What IP addresses do we have?
ip_addresses = [salt.utils.socket_util.IPv4Address(a) for a
in salt.utils.socket_util.ip4_addrs()
if not a.startswith('127.')]
for a in ip_addresses:
if not a.is_private:
log.info('Using public ip address for id: {0}'.format(a))
return str(a), True
if ip_addresses:
a = ip_addresses.pop(0)
log.info('Using private ip address for id: {0}'.format(a))
return str(a), True
log.error('No id found, falling back to localhost')
return 'localhost', False
def launch_server():
"""
Start the salt master here
"""
hostname = get_fqhostname()
ip = host_to_ip(hostname)
log.debug("Hellostack start server %s" % ip)
master_opts = salt.config.client_config("/etc/salt/master")
master = salt.master.Master(master_opts)
master.start()
def _parse_subscriber(self, subscriber):
'''
Parse the subscriber string into the structure needed by _deliver().
'''
log.debug('add recipient: %s', subscriber)
recipient = Recipient(subscriber,
max_msgs=self.max_msgs,
max_age=self.max_age,
state=UNKNOWN,
pending=self.pending)
self.recipients[recipient.addr] = recipient
return recipient
def compound_match(self, tgt):
'''
Runs the compound target check
'''
if not isinstance(tgt, string_types):
log.debug('Compound target received that is not a string')
return False
ref = {'G': 'grain',
'P': 'grain_pcre',
'X': 'exsel',
'I': 'pillar',
'L': 'list',
'S': 'ipcidr',
'E': 'pcre'}
if HAS_RANGE:
ref['R'] = 'range'
results = []
opers = ['and', 'or', 'not', '(', ')']
tokens = re.findall(r'[^\s()]+|[()]', tgt)
for match in tokens:
# Try to match tokens from the compound target, first by using
# the 'G, X, I, L, S, E' matcher types, then by hostname glob.
if '@' in match and match[1] == '@':
comps = match.split('@')
matcher = ref.get(comps[0])
if not matcher:
# If an unknown matcher is called at any time, fail out
return False
results.append(
str(
getattr(self, '{0}_match'.format(matcher))(
'@'.join(comps[1:])
)
)
)
elif match in opers:
# We didn't match a target, so append a boolean operator or
# subexpression
if match == 'not':
if results[-1] == 'and':
pass
elif results[-1] == 'or':
pass
else:
results.append('and')
results.append(match)
else:
# The match is not explicitly defined, evaluate it as a glob
results.append(str(self.glob_match(match)))
return eval(' '.join(results))
def deliver(self, alert):
'''
Deliver an alert sent from a minion.
'''
severity = alert.get('severity')
if severity is not None:
alert['severity'] = severity.lower()
alert['SEVERITY'] = severity.upper()
epoch_time = alert.get('time', time.time())
alert['time'] = time.strftime(self.timeformat,
time.localtime(epoch_time))
alert['verb'] = self.verbs.get(alert.get('verb', DEFAULT_VERB))
log.debug('deliver: %s', alert)
for agent in self.agents.values():
agent.deliver(alert)
def collector(hostname, cmd, result):
'''
Collect data in a mongo database.
'''
conn = pymongo.Connection(
__opts__['mongo.host'],
__opts__['mongo.port'],
)
db = conn[__opts__['mongo.db']]
user = __opts__.get('mongo.user')
password = __opts__.get('mongo.password')
if user and password:
db.authenticate(user, password)
collection = db[hostname]
back = _escape_dot(result)
log.debug(back)
collection.insert({
'utctime': datetime.datetime.utcnow(),
'cmd': cmd,
'result': back
})
def _get_minion_grains(self, *minion_ids, **kwargs):
# Get the minion grains either from cache or from a direct query
# on the minion. By default try to use cached grains first, then
# fall back to querying the minion directly.
ret = {}
cached_grains = kwargs.get('cached_grains', {})
cret = {}
lret = {}
if self.use_cached_grains:
cret = dict([(minion_id, mcache) for (minion_id, mcache) in six.iteritems(cached_grains) if mcache])
missed_minions = [minion_id for minion_id in minion_ids if minion_id not in cret]
log.debug('Missed cached minion grains for: %s', missed_minions)
if self.grains_fallback:
lret = self._get_live_minion_grains(missed_minions)
ret = dict(list(six.iteritems(dict([(minion_id, {}) for minion_id in minion_ids]))) + list(lret.items()) + list(cret.items()))
else:
lret = self._get_live_minion_grains(minion_ids)
missed_minions = [minion_id for minion_id in minion_ids if minion_id not in lret]
log.debug('Missed live minion grains for: %s', missed_minions)
if self.grains_fallback:
cret = dict([(minion_id, mcache) for (minion_id, mcache) in six.iteritems(cached_grains) if mcache])
ret = dict(list(six.iteritems(dict([(minion_id, {}) for minion_id in minion_ids]))) + list(lret.items()) + list(cret.items()))
return ret
def _get_minion_pillar(self, *minion_ids, **kwargs):
# Get the minion pillar either from cache or from a direct query
# on the minion. By default try use the cached pillar first, then
# fall back to rendering pillar on demand with the supplied grains.
ret = {}
grains = kwargs.get('grains', {})
cached_pillar = kwargs.get('cached_pillar', {})
cret = {}
lret = {}
if self.use_cached_pillar:
cret = dict([(minion_id, mcache) for (minion_id, mcache) in cached_pillar.iteritems() if mcache])
missed_minions = [minion_id for minion_id in minion_ids if minion_id not in cret]
log.debug('Missed cached minion pillars for: {0}'.format(missed_minions))
if self.pillar_fallback:
lret = dict([(minion_id, self._get_live_minion_pillar(minion_id, grains.get(minion_id, {}))) for minion_id in missed_minions])
ret = dict(dict([(minion_id, {}) for minion_id in minion_ids]).items() + lret.items() + cret.items())
else:
lret = dict([(minion_id, self._get_live_minion_pillar(minion_id, grains.get(minion_id, {}))) for minion_id in minion_ids])
missed_minions = [minion_id for minion_id in minion_ids if minion_id not in lret]
log.debug('Missed live minion pillars for: {0}'.format(missed_minions))
if self.pillar_fallback:
cret = dict([(minion_id, mcache) for (minion_id, mcache) in cached_pillar.iteritems() if mcache])
ret = dict(dict([(minion_id, {}) for minion_id in minion_ids]).items() + lret.items() + cret.items())
return ret
def _get_minion_pillar(self, *minion_ids, **kwargs):
# Get the minion pillar either from cache or from a direct query
# on the minion. By default try use the cached pillar first, then
# fall back to rendering pillar on demand with the supplied grains.
ret = {}
grains = kwargs.get('grains', {})
cached_pillar = kwargs.get('cached_pillar', {})
cret = {}
lret = {}
if self.use_cached_pillar:
cret = dict([(minion_id, mcache) for (minion_id, mcache) in six.iteritems(cached_pillar) if mcache])
missed_minions = [minion_id for minion_id in minion_ids if minion_id not in cret]
log.debug('Missed cached minion pillars for: %s', missed_minions)
if self.pillar_fallback and missed_minions:
lret = dict([(minion_id, self._get_live_minion_pillar(minion_id, grains.get(minion_id, {}))) for minion_id in missed_minions])
ret = {key: value for key, value in [(minion_id, {}) for minion_id in minion_ids] + list(six.iteritems(cret)) + list(six.iteritems(lret))}
else:
lret = dict([(minion_id, self._get_live_minion_pillar(minion_id, grains.get(minion_id, {}))) for minion_id in minion_ids])
missed_minions = [minion_id for minion_id in minion_ids if minion_id not in lret]
log.debug('Missed live minion pillars for: %s', missed_minions)
if self.pillar_fallback and missed_minions:
cret = dict([(minion_id, mcache) for (minion_id, mcache) in six.iteritems(cached_pillar) if mcache])
ret = {key: value for key, value in [(minion_id, {}) for minion_id in minion_ids] + list(six.iteritems(cret)) + list(six.iteritems(lret))}
return ret
def stop(self, signal, frame):
'''
We override stop() to brake our main loop
and have a pretty log message
'''
log.info("Received signal {0}".format(signal))
# if we have running workers, run through all and join() the ones
# that have finished. if we still have running workers after that,
# wait 5 secs for the rest and then exit. Maybe we should improv
# this a litte bit more
if len(self.running_workers) > 0:
clean_workers = []
for count in range(0, 2):
for worker in self.running_workers:
if worker.isAlive():
clean_workers.append(worker)
else:
worker.join()
log.debug("Joined worker #{0}".format(
worker.getName()))
if len(clean_workers) > 0:
log.info("Waiting 5secs for remaining workers..")
time.sleep(5)
else:
break
log.info("salt-eventsd has shut down")
# leave the cleanup to the supers stop
try:
super(SaltEventsDaemon, self).stop(signal, frame)
except (IOError, OSError):
os._exit(0)
def load_config(opts, path, env_var):
'''
Attempts to update ``opts`` dict by parsing either the file described by
``path`` or the environment variable described by ``env_var`` as YAML.
'''
if path is None:
# When the passed path is None, we just want the configuration
# defaults, not actually loading the whole configuration.
return opts
if not path or not os.path.isfile(path):
path = os.environ.get(env_var, path)
# If the configuration file is missing, attempt to copy the template,
# after removing the first header line.
if not os.path.isfile(path):
template = '{0}.template'.format(path)
if os.path.isfile(template):
import salt.utils # TODO: Need to re-import, need to find out why
with salt.utils.fopen(path, 'w') as out:
with salt.utils.fopen(template, 'r') as f:
f.readline() # skip first line
out.write(f.read())
if os.path.isfile(path):
try:
opts.update(_read_conf_file(path))
opts['conf_file'] = path
except Exception as e:
import salt.log
msg = 'Error parsing configuration file: {0} - {1}'
if salt.log.is_console_configured():
log.warn(msg.format(path, e))
else:
print(msg.format(path, e))
else:
log.debug('Missing configuration file: {0}'.format(path))
def get_minion_pillar(self):
"""
Get pillar data for the targeted minions, either by fetching the
cached minion data on the master, or by compiling the minion's
pillar data on the master.
For runner modules that need access minion pillar data, this
function should be used instead of getting the pillar data by
executing the pillar module on the minions.
By default, this function tries hard to get the pillar data:
- Try to get the cached minion grains and pillar if the
master has minion_data_cache: True
- If the pillar data for the minion is cached, use it.
- If there is no cached grains/pillar data for a minion,
then try to get the minion grains directly from the minion.
- Use the minion grains to compile the pillar directly from the
master using salt.pillar.Pillar
"""
minion_pillars = {}
minion_grains = {}
minion_ids = self._tgt_to_list()
if self.tgt and not minion_ids:
return {}
if any(
arg
for arg in [
self.use_cached_grains,
self.use_cached_pillar,
self.grains_fallback,
self.pillar_fallback,
]
):
log.debug("Getting cached minion data")
cached_minion_grains, cached_minion_pillars = self._get_cached_minion_data(
*minion_ids
)
else:
cached_minion_grains = {}
cached_minion_pillars = {}
log.debug("Getting minion grain data for: %s", minion_ids)
minion_grains = self._get_minion_grains(
*minion_ids, cached_grains=cached_minion_grains
)
log.debug("Getting minion pillar data for: %s", minion_ids)
minion_pillars = self._get_minion_pillar(
*minion_ids, grains=minion_grains, cached_pillar=cached_minion_pillars
)
return minion_pillars
def _read_proc_file(path, opts):
'''
Return a dict of JID metadata, or None
'''
serial = salt.payload.Serial(opts)
with salt.utils.files.fopen(path, 'rb') as fp_:
buf = fp_.read()
fp_.close()
if buf:
data = serial.loads(buf)
else:
# Proc file is empty, remove
try:
os.remove(path)
except IOError:
log.debug('Unable to remove proc file %s.', path)
return None
if not isinstance(data, dict):
# Invalid serial object
return None
if not salt.utils.process.os_is_running(data['pid']):
# The process is no longer running, clear out the file and
# continue
try:
os.remove(path)
except IOError:
log.debug('Unable to remove proc file %s.', path)
return None
if not _check_cmdline(data):
pid = data.get('pid')
if pid:
log.warning(
'PID %s exists but does not appear to be a salt process.', pid
)
try:
os.remove(path)
except IOError:
log.debug('Unable to remove proc file %s.', path)
return None
return data
A single monitor task.
'''
def __init__(self, taskid, pyexe, context, scheduler=None):
self.taskid = taskid
self.code = pyexe
self.context = context
self.scheduler = scheduler
def run(self):
log.trace('start thread for %s', self.taskid)
minion = self.context.get('id')
collector = self.context.get('collector')
while True:
try:
exec self.code in self.context
except Exception, ex:
log.error("can't execute %s: %s", self.taskid, ex, exc_info=ex)
if collector:
jid = datetime.datetime.strftime(
datetime.datetime.now(), 'M%Y%m%d%H%M%S%f')
try:
collector(minion, self.context['cmd'], self.context['result'])
except Exception, ex:
log.error('monitor error: %s', self.taskid, exc_info=ex)
if self.scheduler is None:
break
duration = self.scheduler.next()
log.trace('%s: sleep %s seconds', self.taskid, duration)
time.sleep(duration)
log.debug('thread exit: %s', self.taskid)
def loads(self, msg, encoding=None, raw=False):
"""
Run the correct loads serialization format
:param encoding: Useful for Python 3 support. If the msgpack data
was encoded using "use_bin_type=True", this will
differentiate between the 'bytes' type and the
'str' type by decoding contents with 'str' type
to what the encoding was set as. Recommended
encoding is 'utf-8' when using Python 3.
If the msgpack data was not encoded using
"use_bin_type=True", it will try to decode
all 'bytes' and 'str' data (the distinction has
been lost in this case) to what the encoding is
set as. In this case, it will fail if any of
the contents cannot be converted.
"""
try:
def ext_type_decoder(code, data):
if code == 78:
data = salt.utils.stringutils.to_unicode(data)
return datetime.datetime.strptime(data,
"%Y%m%dT%H:%M:%S.%f")
return data
gc.disable() # performance optimization for msgpack
loads_kwargs = {"use_list": True, "ext_hook": ext_type_decoder}
if salt.utils.msgpack.version >= (0, 4, 0):
# msgpack only supports 'encoding' starting in 0.4.0.
# Due to this, if we don't need it, don't pass it at all so
# that under Python 2 we can still work with older versions
# of msgpack.
if salt.utils.msgpack.version >= (0, 5, 2):
if encoding is None:
loads_kwargs["raw"] = True
else:
loads_kwargs["raw"] = False
else:
loads_kwargs["encoding"] = encoding
try:
ret = salt.utils.msgpack.unpackb(msg, **loads_kwargs)
except UnicodeDecodeError:
# msg contains binary data
loads_kwargs.pop("raw", None)
loads_kwargs.pop("encoding", None)
ret = salt.utils.msgpack.loads(msg, **loads_kwargs)
else:
ret = salt.utils.msgpack.loads(msg, **loads_kwargs)
if encoding is None and not raw:
ret = salt.transport.frame.decode_embedded_strs(ret)
except Exception as exc: # pylint: disable=broad-except
log.critical(
"Could not deserialize msgpack message. This often happens "
"when trying to read a file not in binary mode. "
"To see message payload, enable debug logging and retry. "
"Exception: %s",
exc,
)
log.debug("Msgpack deserialization failure on message: %s", msg)
gc.collect()
exc_msg = "Could not deserialize msgpack message. See log for more info."
raise SaltDeserializationError(exc_msg) from exc
finally:
gc.enable()
return ret
def _thread_return(class_, minion_instance, opts, data):
'''
This method should be used as a threading target, start the actual
minion side execution.
'''
# this seems awkward at first, but it's a workaround for Windows
# multiprocessing communication.
if not minion_instance:
minion_instance = class_(opts)
if opts['multiprocessing']:
fn_ = os.path.join(minion_instance.proc_dir, data['jid'])
sdata = {'pid': os.getpid()}
sdata.update(data)
with salt.utils.fopen(fn_, 'w+') as fp_:
fp_.write(minion_instance.serial.dumps(sdata))
ret = {}
for ind in range(0, len(data['arg'])):
try:
arg = eval(data['arg'][ind])
if isinstance(arg, bool):
data['arg'][ind] = str(data['arg'][ind])
elif isinstance(arg, (dict, int, list, string_types)):
data['arg'][ind] = arg
else:
data['arg'][ind] = str(data['arg'][ind])
except Exception:
pass
function_name = data['fun']
if function_name in minion_instance.functions:
ret['success'] = False
try:
func = minion_instance.functions[data['fun']]
args, kwargs = detect_kwargs(func, data['arg'], data)
ret['return'] = func(*args, **kwargs)
ret['success'] = True
except CommandNotFoundError as exc:
msg = 'Command required for \'{0}\' not found: {1}'
log.debug(msg.format(function_name, str(exc)))
ret['return'] = msg.format(function_name, str(exc))
except CommandExecutionError as exc:
msg = 'A command in {0} had a problem: {1}'
log.error(msg.format(function_name, str(exc)))
ret['return'] = 'ERROR: {0}'.format(str(exc))
except SaltInvocationError as exc:
msg = 'Problem executing "{0}": {1}'
log.error(msg.format(function_name, str(exc)))
ret['return'] = 'ERROR executing {0}: {1}'.format(
function_name, exc)
except Exception:
trb = traceback.format_exc()
msg = 'The minion function caused an exception: {0}'
log.warning(msg.format(trb))
ret['return'] = trb
else:
ret['return'] = '"{0}" is not available.'.format(function_name)
ret['jid'] = data['jid']
ret['fun'] = data['fun']
minion_instance._return_pub(ret)
if data['ret']:
for returner in set(data['ret'].split(',')):
ret['id'] = opts['id']
try:
minion_instance.returners['{0}.returner'.format(returner)](
ret)
except Exception as exc:
log.error('The return failed for job {0} {1}'.format(
data['jid'], exc))
def _linux_gpu_data():
'''
num_gpus: int
gpus:
- vendor: nvidia|amd|ati|...
model: string
'''
lspci = salt.utils.which('lspci')
if not lspci:
log.info(
'The `lspci` binary is not available on the system. GPU grains '
'will not be available.')
return {}
elif __opts__.get('enable_gpu_grains', None) is False:
log.info(
'Skipping lspci call because enable_gpu_grains was set to False '
'in the config. GPU grains will not be available.')
return {}
# dominant gpu vendors to search for (MUST be lowercase for matching below)
known_vendors = ['nvidia', 'amd', 'ati', 'intel']
devs = []
try:
lspci_out = __salt__['cmd.run']('lspci -vmm')
cur_dev = {}
error = False
# Add a blank element to the lspci_out.splitlines() list,
# otherwise the last device is not evaluated as a cur_dev and ignored.
lspci_list = lspci_out.splitlines()
lspci_list.append('')
for line in lspci_list:
# check for record-separating empty lines
if line == '':
if cur_dev.get('Class', '') == 'VGA compatible controller':
devs.append(cur_dev)
# XXX; may also need to search for "3D controller"
cur_dev = {}
continue
if re.match(r'^\w+:\s+.*', line):
key, val = line.split(':', 1)
cur_dev[key.strip()] = val.strip()
else:
error = True
log.debug('Unexpected lspci output: \'{0}\''.format(line))
if error:
log.warn('Error loading grains, unexpected linux_gpu_data output, '
'check that you have a valid shell configured and '
'permissions to run lspci command')
except OSError:
pass
gpus = []
for gpu in devs:
vendor_strings = gpu['Vendor'].lower().split()
# default vendor to 'unknown', overwrite if we match a known one
vendor = 'unknown'
for name in known_vendors:
# search for an 'expected' vendor name in the list of strings
if name in vendor_strings:
vendor = name
break
gpus.append({'vendor': vendor, 'model': gpu['Device']})
grains = {}
grains['num_gpus'] = len(gpus)
grains['gpus'] = gpus
return grains
def cmd_block(self, is_retry=False):
'''
Prepare the pre-check command to send to the subsystem
'''
# 1. execute SHIM + command
# 2. check if SHIM returns a master request or if it completed
# 3. handle any master request
# 4. re-execute SHIM + command
# 5. split SHIM results from command results
# 6. return command results
log.debug('Performing shimmed, blocking command as follows:\n{0}'.format(' '.join(self.argv)))
cmd_str = self._cmd_str()
stdout, stderr, retcode = self.shell.exec_cmd(cmd_str)
log.debug('STDOUT {1}\n{0}'.format(stdout, self.target['host']))
log.debug('STDERR {1}\n{0}'.format(stderr, self.target['host']))
log.debug('RETCODE {1}: {0}'.format(retcode, self.target['host']))
error = self.categorize_shim_errors(stdout, stderr, retcode)
if error:
if error == 'Undefined SHIM state':
self.deploy()
stdout, stderr, retcode = self.shell.exec_cmd(cmd_str)
if not re.search(RSTR_RE, stdout) or not re.search(RSTR_RE, stderr):
# If RSTR is not seen in both stdout and stderr then there
# was a thin deployment problem.
return 'ERROR: Failure deploying thin: {0}'.format(stdout), stderr, retcode
stdout = re.split(RSTR_RE, stdout, 1)[1].strip()
stderr = re.split(RSTR_RE, stderr, 1)[1].strip()
else:
return 'ERROR: {0}'.format(error), stderr, retcode
# FIXME: this discards output from ssh_shim if the shim succeeds. It should
# always save the shim output regardless of shim success or failure.
if re.search(RSTR_RE, stdout):
stdout = re.split(RSTR_RE, stdout, 1)[1].strip()
else:
# This is actually an error state prior to the shim but let it fall through
pass
if re.search(RSTR_RE, stderr):
# Found RSTR in stderr which means SHIM completed and only
# and remaining output is only from salt.
stderr = re.split(RSTR_RE, stderr, 1)[1].strip()
else:
# RSTR was found in stdout but not stderr - which means there
# is a SHIM command for the master.
shim_command = re.split(r'\r?\n', stdout, 1)[0].strip()
if 'deploy' == shim_command and retcode == salt.exitcodes.EX_THIN_DEPLOY:
self.deploy()
stdout, stderr, retcode = self.shell.exec_cmd(cmd_str)
if not re.search(RSTR_RE, stdout) or not re.search(RSTR_RE, stderr):
# If RSTR is not seen in both stdout and stderr then there
# was a thin deployment problem.
return 'ERROR: Failure deploying thin: {0}'.format(stdout), stderr, retcode
stdout = re.split(RSTR_RE, stdout, 1)[1].strip()
stderr = re.split(RSTR_RE, stderr, 1)[1].strip()
return stdout, stderr, retcode
def load_config(path, env_var, default_path=None):
'''
Returns configuration dict from parsing either the file described by
``path`` or the environment variable described by ``env_var`` as YAML.
'''
if path is None:
# When the passed path is None, we just want the configuration
# defaults, not actually loading the whole configuration.
return {}
if default_path is None:
# This is most likely not being used from salt, ie, could be salt-cloud
# or salt-api which have not yet migrated to the new default_path
# argument. Let's issue a warning message that the environ vars won't
# work.
import inspect
previous_frame = inspect.getframeinfo(inspect.currentframe().f_back)
log.warning(
'The function \'{0}()\' defined in {1!r} is not yet using the '
'new \'default_path\' argument to `salt.config.load_config()`. '
'As such, the {2!r} environment variable will be ignored'.format(
previous_frame.function, previous_frame.filename, env_var))
# In this case, maintain old behaviour
default_path = DEFAULT_MASTER_OPTS['conf_file']
# Default to the environment variable path, if it exists
env_path = os.environ.get(env_var, path)
if not env_path or not os.path.isfile(env_path):
env_path = path
# If non-default path from `-c`, use that over the env variable
if path != default_path:
env_path = path
path = env_path
# If the configuration file is missing, attempt to copy the template,
# after removing the first header line.
if not os.path.isfile(path):
template = '{0}.template'.format(path)
if os.path.isfile(template):
import salt.utils # TODO: Need to re-import, need to find out why
log.debug('Writing {0} based on {1}'.format(path, template))
with salt.utils.fopen(path, 'w') as out:
with salt.utils.fopen(template, 'r') as ifile:
ifile.readline() # skip first line
out.write(ifile.read())
if os.path.isfile(path):
try:
opts = _read_conf_file(path)
opts['conf_file'] = path
return opts
except Exception as err:
import salt.log
msg = 'Error parsing configuration file: {0} - {1}'
if salt.log.is_console_configured():
log.warn(msg.format(path, err))
else:
print(msg.format(path, err))
else:
log.debug('Missing configuration file: {0}'.format(path))
return {}
def run(self):
'''
Main loop of the FSCache, checks schedule, retrieves result-data
from the workers and answer requests with data from the cache
'''
context = zmq.Context()
# the socket for incoming cache requests
creq_in = context.socket(zmq.REP)
creq_in.setsockopt(zmq.LINGER, 100)
creq_in.bind('ipc:///' + self.cache_sock)
# the socket for incoming cache-updates from workers
cupd_in = context.socket(zmq.REP)
cupd_in.setsockopt(zmq.LINGER, 100)
cupd_in.bind('ipc:///' + self.update_sock)
# wait for the timer to bind to its socket
log.debug('wait 2 secs for the timer')
time.sleep(2)
# the socket for the timer-event
timer_in = context.socket(zmq.PULL)
timer_in.setsockopt(zmq.LINGER, 100)
timer_in.connect('ipc:///' + self.upd_t_sock)
poller = zmq.Poller()
poller.register(creq_in, zmq.POLLIN)
poller.register(cupd_in, zmq.POLLIN)
poller.register(timer_in, zmq.POLLIN)
# our serializer
serial = salt.payload.Serial(self.opts.get('serial', ''))
# register a signal handler
signal.signal(signal.SIGINT, self.signal_handler)
# secure the sockets from the world
self.secure()
log.info('FSCache started')
log.debug('FSCache started')
while self.running:
# we check for new events with the poller
try:
socks = dict(poller.poll())
except KeyboardInterrupt:
self.stop()
except zmq.ZMQError as t:
self.stop()
# check for next cache-request
if socks.get(creq_in) == zmq.POLLIN:
msg = serial.loads(creq_in.recv())
log.debug('Received request: {0}'.format(msg))
# we only accept requests as lists [req_id, <path>]
if isinstance(msg, list):
# for now only one item is assumed to be requested
msgid, file_n = msg[:]
log.debug('Looking for {0}:{1}'.format(msgid, file_n))
fdata = self.path_data.get(file_n, None)
if fdata is not None:
log.debug('Cache HIT')
else:
log.debug('Cache MISS')
# simulate slow caches
#randsleep = random.randint(0,3)
#time.sleep(randsleep)
# Send reply back to client
reply = serial.dumps([msgid, fdata])
creq_in.send(reply)
# wrong format, item not cached
else:
reply = serial.dumps([msgid, None])
creq_in.send(reply)
# check for next cache-update from workers
elif socks.get(cupd_in) == zmq.POLLIN:
new_c_data = serial.loads(cupd_in.recv())
# tell the worker to exit
cupd_in.send(serial.dumps('OK'))
# check if the returned data is usable
if not isinstance(new_c_data, dict):
log.error('Worker returned unusable result')
del new_c_data
continue
# the workers will return differing data:
# 1. '{'file1': <data1>, 'file2': <data2>,...}' - a cache update
# 2. '{search-path: None}' - job was not run, pre-checks failed
# 3. '{}' - no files found, check the pattern if defined?
# 4. anything else is considered malformed
if len(new_c_data) == 0:
log.debug('Got empty update from worker')
elif new_c_data.values()[0] is not None:
log.debug('Got cache update with {0} item(s)'.format(
len(new_c_data)))
self.path_data.update(new_c_data)
else:
log.debug('Got malformed result dict from worker')
log.info('{0} entries in cache'.format(len(self.path_data)))
# check for next timer-event to start new jobs
elif socks.get(timer_in) == zmq.POLLIN:
sec_event = serial.loads(timer_in.recv())
log.debug('Timer event: #{0}'.format(sec_event))
# loop through the jobs and start if a jobs ival matches
for item in self.jobs:
if sec_event in self.jobs[item]['ival']:
self.run_job(item)
self.stop()
creq_in.close()
cupd_in.close()
timer_in.close()
context.term()
log.debug('Shutting down')\
def listen(self):
'''
the main event loop where we receive the events and
start the workers that dump our data into the database
'''
# log on to saltstacks event-bus
event = salt.utils.event.SaltEvent(self.node,
self.sock_dir)
# we store our events in a list, we dont really care about an order
# or what kind of data is put in there. all that is configured with the
# sql-template configured in the configfile
event_queue = []
# start our dump_timer
self.ev_timer.start()
# this is for logline chronology so the timer-message always comes
# _before_ the actual startup-message of the listening loop below :-)
time.sleep(1)
log.info("entering main event loop")
log.info("listening on: {0}".format(event.puburi))
# read everything we can get our hands on
while True:
# the zmq-socket does not like ^C very much, make the error
# a little more graceful. alright, alright, ignore the damn thing,
# we're exiting anyways...
try:
ret = event.get_event(full=True)
except zmq.ZMQError:
pass
if ret is None:
continue
# if the timer has expired, we may have not received enough
# events in the queue to reach event_limit, in that case we dump
# the data anyway to have it in the database
if(self.ev_timer_ev):
if (len(self.running_workers) < self.max_workers) and \
(len(event_queue) > 0):
self._init_worker(event_queue)
# reset our queue to prevent duplicate entries
del event_queue[:]
# we reset the timer.ev_timer_ev at the end of the loop
# so we can update the stats that are logged
# filter only the events we're interested in. all events have a tag
# we can filter them by. we match with a precompiled regex
if( 'tag' in ret ):
# filter out events with an empty tag. those are special
if( ret['tag'] != '' ):
# run through our configured events and try to match the
# current events tag against the ones we're interested in
for key in self.event_map.keys():
if( self.event_map[key]['tag'].match(ret['tag'])):
log.debug("matching on {0}:{1}".format(key,
ret['tag']))
prio = self.event_map[key].get('prio', 0)
# push prio1-events directly into a worker
if prio > 0:
log.debug('Prio1 event found, pushing immediately!')
self.events_han += 1
self._init_worker([ret])
else:
event_queue.append(ret)
self.events_han += 1
# once we reach the event_limit, start a worker that
# writes that data in to the database
if len(event_queue) >= self.event_limit:
# only start a worker if not too many workers are running
if len(self.running_workers) < self.max_workers:
self._init_worker(event_queue)
# reset the timer
self.ev_timer.reset()
# reset our queue to prevent duplicate entries
del event_queue[:]
else:
# FIXME: we need to handle this situation somehow if
# too many workers are running. just flush the events?
# there really is no sane way except queueing more and more
# until some sort of limit is reached and we care more about
# our saltmaster than about the collected events!
log.critical("too many workers running, loosing data!!!")
# a list for the workers that are still running
clean_workers = []
# run through all the workers and join() the ones
# that have finished dumping their data and keep
# the running ones on our list
for worker in self.running_workers:
if worker.isAlive():
clean_workers.append(worker)
else:
worker.join()
log.debug("joined worker #{0}".format(worker.getName()))
self.threads_join += 1
# get rid of the old reference and set a new one
# FIXME: is this really neccessary?
del self.running_workers
self.running_workers = clean_workers
self.events_rec += 1
# we update the stats every 'received div handled == 0'
# or if we recevied a timer event from our ResetTimer
if( (self.events_rec % self.state_upd) == 0 ):
self._write_state()
elif(self.ev_timer_ev):
self._write_state()
self.ev_timer_ev = False
log.info("listen loop ended...")
def loads(self, msg, encoding=None, raw=False):
'''
Run the correct loads serialization format
:param encoding: Useful for Python 3 support. If the msgpack data
was encoded using "use_bin_type=True", this will
differentiate between the 'bytes' type and the
'str' type by decoding contents with 'str' type
to what the encoding was set as. Recommended
encoding is 'utf-8' when using Python 3.
If the msgpack data was not encoded using
"use_bin_type=True", it will try to decode
all 'bytes' and 'str' data (the distinction has
been lost in this case) to what the encoding is
set as. In this case, it will fail if any of
the contents cannot be converted.
'''
try:
def ext_type_decoder(code, data):
if code == 78:
data = salt.utils.stringutils.to_unicode(data)
return datetime.datetime.strptime(data,
'%Y%m%dT%H:%M:%S.%f')
return data
gc.disable() # performance optimization for msgpack
if msgpack.version >= (0, 4, 0):
# msgpack only supports 'encoding' starting in 0.4.0.
# Due to this, if we don't need it, don't pass it at all so
# that under Python 2 we can still work with older versions
# of msgpack.
try:
ret = salt.utils.msgpack.loads(msg,
use_list=True,
ext_hook=ext_type_decoder,
encoding=encoding,
_msgpack_module=msgpack)
except UnicodeDecodeError:
# msg contains binary data
ret = msgpack.loads(msg,
use_list=True,
ext_hook=ext_type_decoder)
else:
ret = salt.utils.msgpack.loads(msg,
use_list=True,
ext_hook=ext_type_decoder,
_msgpack_module=msgpack)
if six.PY3 and encoding is None and not raw:
ret = salt.transport.frame.decode_embedded_strs(ret)
except Exception as exc:
log.critical(
'Could not deserialize msgpack message. This often happens '
'when trying to read a file not in binary mode. '
'To see message payload, enable debug logging and retry. '
'Exception: %s', exc)
log.debug('Msgpack deserialization failure on message: %s', msg)
gc.collect()
raise
finally:
gc.enable()
return ret
def cmd_block(self, is_retry=False):
'''
Prepare the pre-check command to send to the subsystem
1. execute SHIM + command
2. check if SHIM returns a master request or if it completed
3. handle any master request
4. re-execute SHIM + command
5. split SHIM results from command results
6. return command results
'''
self.argv = _convert_args(self.argv)
log.debug('Performing shimmed, blocking command as follows:\n{0}'.format(' '.join(self.argv)))
cmd_str = self._cmd_str()
stdout, stderr, retcode = self.shim_cmd(cmd_str)
log.trace('STDOUT {1}\n{0}'.format(stdout, self.target['host']))
log.trace('STDERR {1}\n{0}'.format(stderr, self.target['host']))
log.debug('RETCODE {1}: {0}'.format(retcode, self.target['host']))
error = self.categorize_shim_errors(stdout, stderr, retcode)
if error:
if error == 'Undefined SHIM state':
self.deploy()
stdout, stderr, retcode = self.shim_cmd(cmd_str)
if not re.search(RSTR_RE, stdout) or not re.search(RSTR_RE, stderr):
# If RSTR is not seen in both stdout and stderr then there
# was a thin deployment problem.
return 'ERROR: Failure deploying thin, undefined state: {0}'.format(stdout), stderr, retcode
while re.search(RSTR_RE, stdout):
stdout = re.split(RSTR_RE, stdout, 1)[1].strip()
while re.search(RSTR_RE, stderr):
stderr = re.split(RSTR_RE, stderr, 1)[1].strip()
else:
return 'ERROR: {0}'.format(error), stderr, retcode
# FIXME: this discards output from ssh_shim if the shim succeeds. It should
# always save the shim output regardless of shim success or failure.
while re.search(RSTR_RE, stdout):
stdout = re.split(RSTR_RE, stdout, 1)[1].strip()
if re.search(RSTR_RE, stderr):
# Found RSTR in stderr which means SHIM completed and only
# and remaining output is only from salt.
while re.search(RSTR_RE, stderr):
stderr = re.split(RSTR_RE, stderr, 1)[1].strip()
else:
# RSTR was found in stdout but not stderr - which means there
# is a SHIM command for the master.
shim_command = re.split(r'\r?\n', stdout, 1)[0].strip()
log.debug('SHIM retcode({0}) and command: {1}'.format(retcode, shim_command))
if 'deploy' == shim_command and retcode == salt.defaults.exitcodes.EX_THIN_DEPLOY:
self.deploy()
stdout, stderr, retcode = self.shim_cmd(cmd_str)
if not re.search(RSTR_RE, stdout) or not re.search(RSTR_RE, stderr):
if not self.tty:
# If RSTR is not seen in both stdout and stderr then there
# was a thin deployment problem.
log.error('ERROR: Failure deploying thin, retrying: {0}\n{1}'.format(stdout, stderr), stderr, retcode)
return self.cmd_block()
elif not re.search(RSTR_RE, stdout):
# If RSTR is not seen in stdout with tty, then there
# was a thin deployment problem.
log.error('ERROR: Failure deploying thin, retrying: {0}\n{1}'.format(stdout, stderr), stderr, retcode)
while re.search(RSTR_RE, stdout):
stdout = re.split(RSTR_RE, stdout, 1)[1].strip()
if self.tty:
stderr = ''
else:
while re.search(RSTR_RE, stderr):
stderr = re.split(RSTR_RE, stderr, 1)[1].strip()
elif 'ext_mods' == shim_command:
self.deploy_ext()
stdout, stderr, retcode = self.shim_cmd(cmd_str)
if not re.search(RSTR_RE, stdout) or not re.search(RSTR_RE, stderr):
# If RSTR is not seen in both stdout and stderr then there
# was a thin deployment problem.
return 'ERROR: Failure deploying ext_mods: {0}'.format(stdout), stderr, retcode
while re.search(RSTR_RE, stdout):
stdout = re.split(RSTR_RE, stdout, 1)[1].strip()
while re.search(RSTR_RE, stderr):
stderr = re.split(RSTR_RE, stderr, 1)[1].strip()
return stdout, stderr, retcode
def tune_in(self):
'''
Lock onto the publisher. This is the main event loop for the minion
'''
log.info('{0} is starting as user \'{1}\''.format(
self.__class__.__name__, getpass.getuser()))
log.debug('Minion "{0}" trying to tune in'.format(self.opts['id']))
self.context = zmq.Context()
# Prepare the minion event system
#
# Start with the publish socket
id_hash = hashlib.md5(self.opts['id']).hexdigest()
epub_sock_path = os.path.join(
self.opts['sock_dir'], 'minion_event_{0}_pub.ipc'.format(id_hash))
epull_sock_path = os.path.join(
self.opts['sock_dir'], 'minion_event_{0}_pull.ipc'.format(id_hash))
self.epub_sock = self.context.socket(zmq.PUB)
if self.opts.get('ipc_mode', '') == 'tcp':
epub_uri = 'tcp://127.0.0.1:{0}'.format(self.opts['tcp_pub_port'])
epull_uri = 'tcp://127.0.0.1:{0}'.format(
self.opts['tcp_pull_port'])
else:
epub_uri = 'ipc://{0}'.format(epub_sock_path)
salt.utils.check_ipc_path_max_len(epub_uri)
epull_uri = 'ipc://{0}'.format(epull_sock_path)
salt.utils.check_ipc_path_max_len(epull_uri)
log.debug('{0} PUB socket URI: {1}'.format(self.__class__.__name__,
epub_uri))
log.debug('{0} PULL socket URI: {1}'.format(self.__class__.__name__,
epull_uri))
# Create the pull socket
self.epull_sock = self.context.socket(zmq.PULL)
# Bind the event sockets
self.epub_sock.bind(epub_uri)
self.epull_sock.bind(epull_uri)
# Restrict access to the sockets
if not self.opts.get('ipc_mode', '') == 'tcp':
os.chmod(epub_sock_path, 448)
os.chmod(epull_sock_path, 448)
self.poller = zmq.Poller()
self.epoller = zmq.Poller()
self.socket = self.context.socket(zmq.SUB)
self.socket.setsockopt(zmq.SUBSCRIBE, '')
self.socket.setsockopt(zmq.IDENTITY, self.opts['id'])
if hasattr(zmq, 'RECONNECT_IVL_MAX'):
self.socket.setsockopt(zmq.RECONNECT_IVL_MAX,
self.opts['recon_max'])
if hasattr(zmq, 'TCP_KEEPALIVE'):
self.socket.setsockopt(zmq.TCP_KEEPALIVE,
self.opts['tcp_keepalive'])
self.socket.setsockopt(zmq.TCP_KEEPALIVE_IDLE,
self.opts['tcp_keepalive_idle'])
self.socket.setsockopt(zmq.TCP_KEEPALIVE_CNT,
self.opts['tcp_keepalive_cnt'])
self.socket.setsockopt(zmq.TCP_KEEPALIVE_INTVL,
self.opts['tcp_keepalive_intvl'])
if hasattr(zmq, 'IPV4ONLY'):
self.socket.setsockopt(
zmq.IPV4ONLY,
int(not int(self.opts.get('ipv6_enable', False))))
self.socket.connect(self.master_pub)
self.poller.register(self.socket, zmq.POLLIN)
self.epoller.register(self.epull_sock, zmq.POLLIN)
# Send an event to the master that the minion is live
self._fire_master(
'Minion {0} started at {1}'.format(self.opts['id'],
time.asctime()), 'minion_start')
if self.opts['multiprocessing'] and not salt.utils.is_windows():
signal.signal(signal.SIGCHLD, self.handle_sigchld)
# Make sure to gracefully handle SIGUSR1
enable_sigusr1_handler()
# On first startup execute a state run if configured to do so
self._state_run()
while True:
try:
self.schedule.eval()
socks = dict(
self.poller.poll(self.opts['loop_interval'] * 1000))
if self.socket in socks and socks[self.socket] == zmq.POLLIN:
payload = self.serial.loads(self.socket.recv())
self._handle_payload(payload)
time.sleep(0.05)
# Clean up the minion processes which have been executed and
# have finished
# Check if modules and grains need to be refreshed
self.passive_refresh()
# Check the event system
if self.epoller.poll(1):
try:
package = self.epull_sock.recv(zmq.NOBLOCK)
self.epub_sock.send(package)
except Exception:
pass
except zmq.ZMQError:
# This is thrown by the inturupt caused by python handling the
# SIGCHLD. This is a safe error and we just start the poll
# again
continue
except Exception:
log.critical(traceback.format_exc())
def clear_cached_minion_data(self,
clear_pillar=False,
clear_grains=False,
clear_mine=False,
clear_mine_func=None):
'''
Clear the cached data/files for the targeted minions.
'''
clear_what = []
if clear_pillar:
clear_what.append('pillar')
if clear_grains:
clear_what.append('grains')
if clear_mine:
clear_what.append('mine')
if clear_mine_func is not None:
clear_what.append('mine_func: \'{0}\''.format(clear_mine_func))
if not clear_what:
log.debug('No cached data types specified for clearing.')
return False
minion_ids = self._tgt_to_list()
log.debug('Clearing cached %s data for: %s', ', '.join(clear_what),
minion_ids)
if clear_pillar == clear_grains:
# clear_pillar and clear_grains are both True or both False.
# This means we don't deal with pillar/grains caches at all.
grains = {}
pillars = {}
else:
# Unless both clear_pillar and clear_grains are True, we need
# to read in the pillar/grains data since they are both stored
# in the same file, 'data.p'
grains, pillars = self._get_cached_minion_data(*minion_ids)
try:
c_minions = self.cache.list('minions')
for minion_id in minion_ids:
if not salt.utils.verify.valid_id(self.opts, minion_id):
continue
if minion_id not in c_minions:
# Cache bank for this minion does not exist. Nothing to do.
continue
bank = 'minions/{0}'.format(minion_id)
minion_pillar = pillars.pop(minion_id, False)
minion_grains = grains.pop(minion_id, False)
if ((clear_pillar and clear_grains)
or (clear_pillar and not minion_grains)
or (clear_grains and not minion_pillar)):
# Not saving pillar or grains, so just delete the cache file
self.cache.flush(bank, 'data')
elif clear_pillar and minion_grains:
self.cache.store(bank, 'data', {'grains': minion_grains})
elif clear_grains and minion_pillar:
self.cache.store(bank, 'data', {'pillar': minion_pillar})
if clear_mine:
# Delete the whole mine file
self.cache.flush(bank, 'mine')
elif clear_mine_func is not None:
# Delete a specific function from the mine file
mine_data = self.cache.fetch(bank, 'mine')
if isinstance(mine_data, dict):
if mine_data.pop(clear_mine_func, False):
self.cache.store(bank, 'mine', mine_data)
except (OSError, IOError):
return True
return True
def clear_cached_minion_data(self,
clear_pillar=False,
clear_grains=False,
clear_mine=False,
clear_mine_func=None):
'''
Clear the cached data/files for the targeted minions.
'''
clear_what = []
if clear_pillar:
clear_what.append('pillar')
if clear_grains:
clear_what.append('grains')
if clear_mine:
clear_what.append('mine')
if clear_mine_func is not None:
clear_what.append('mine_func: {0!r}'.format(clear_mine_func))
if not len(clear_what):
log.debug('No cached data types specified for clearing.')
return False
minion_ids = self._tgt_to_list()
log.debug('Clearing cached {0} data for: {1}'.format(
', '.join(clear_what),
minion_ids))
if clear_pillar == clear_grains:
# clear_pillar and clear_grains are both True or both False.
# This means we don't deal with pillar/grains caches at all.
grains = {}
pillars = {}
else:
# Unless both clear_pillar and clear_grains are True, we need
# to read in the pillar/grains data since they are both stored
# in the same file, 'data.p'
grains, pillars = self._get_cached_minion_data(*minion_ids)
try:
for minion_id in minion_ids:
if not salt.utils.verify.valid_id(self.opts, minion_id):
continue
cdir = os.path.join(self.opts['cachedir'], 'minions', minion_id)
if not os.path.isdir(cdir):
# Cache dir for this minion does not exist. Nothing to do.
continue
data_file = os.path.join(cdir, 'data.p')
mine_file = os.path.join(cdir, 'mine.p')
minion_pillar = pillars.pop(minion_id, False)
minion_grains = grains.pop(minion_id, False)
if ((clear_pillar and clear_grains) or
(clear_pillar and not minion_grains) or
(clear_grains and not minion_pillar)):
# Not saving pillar or grains, so just delete the cache file
os.remove(os.path.join(data_file))
elif clear_pillar and minion_grains:
tmpfh, tmpfname = tempfile.mkstemp(dir=cdir)
os.close(tmpfh)
with salt.utils.fopen(tmpfname, 'w+b') as fp_:
fp_.write(self.serial.dumps({'grains': minion_grains}))
os.rename(tmpfname, data_file)
elif clear_grains and minion_pillar:
tmpfh, tmpfname = tempfile.mkstemp(dir=cdir)
os.close(tmpfh)
with salt.utils.fopen(tmpfname, 'w+b') as fp_:
fp_.write(self.serial.dumps({'pillar': minion_pillar}))
os.rename(tmpfname, data_file)
if clear_mine:
# Delete the whole mine file
os.remove(os.path.join(mine_file))
elif clear_mine_func is not None:
# Delete a specific function from the mine file
with salt.utils.fopen(mine_file, 'rb') as fp_:
mine_data = self.serial.loads(fp_.read())
if isinstance(mine_data, dict):
if mine_data.pop(clear_mine_func, False):
tmpfh, tmpfname = tempfile.mkstemp(dir=cdir)
os.close(tmpfh)
with salt.utils.fopen(tmpfname, 'w+b') as fp_:
fp_.write(self.serial.dumps(mine_data))
os.rename(tmpfname, mine_file)
except (OSError, IOError):
return True
return True
def run(self):
'''
Main loop of the ConCache, starts updates in intervals and
answers requests from the MWorkers
'''
context = zmq.Context()
# the socket for incoming cache requests
creq_in = context.socket(zmq.REP)
creq_in.setsockopt(zmq.LINGER, 100)
creq_in.bind('ipc://' + self.cache_sock)
# the socket for incoming cache-updates from workers
cupd_in = context.socket(zmq.SUB)
cupd_in.setsockopt(zmq.SUBSCRIBE, b'')
cupd_in.setsockopt(zmq.LINGER, 100)
cupd_in.bind('ipc://' + self.update_sock)
# the socket for the timer-event
timer_in = context.socket(zmq.SUB)
timer_in.setsockopt(zmq.SUBSCRIBE, b'')
timer_in.setsockopt(zmq.LINGER, 100)
timer_in.connect('ipc://' + self.upd_t_sock)
poller = zmq.Poller()
poller.register(creq_in, zmq.POLLIN)
poller.register(cupd_in, zmq.POLLIN)
poller.register(timer_in, zmq.POLLIN)
# our serializer
serial = salt.payload.Serial(self.opts.get('serial', ''))
# register a signal handler
signal.signal(signal.SIGINT, self.signal_handler)
# secure the sockets from the world
self.secure()
log.info('ConCache started')
while self.running:
# we check for new events with the poller
try:
socks = dict(poller.poll(1))
except KeyboardInterrupt:
self.stop()
except zmq.ZMQError as zmq_err:
log.error('ConCache ZeroMQ-Error occurred')
log.exception(zmq_err)
self.stop()
# check for next cache-request
if socks.get(creq_in) == zmq.POLLIN:
msg = serial.loads(creq_in.recv())
log.debug('ConCache Received request: %s', msg)
# requests to the minion list are send as str's
if isinstance(msg, six.string_types):
if msg == 'minions':
# Send reply back to client
reply = serial.dumps(self.minions)
creq_in.send(reply)
# check for next cache-update from workers
if socks.get(cupd_in) == zmq.POLLIN:
new_c_data = serial.loads(cupd_in.recv())
# tell the worker to exit
#cupd_in.send(serial.dumps('ACK'))
# check if the returned data is usable
if not isinstance(new_c_data, list):
log.error('ConCache Worker returned unusable result')
del new_c_data
continue
# the cache will receive lists of minions
# 1. if the list only has 1 item, its from an MWorker, we append it
# 2. if the list contains another list, its from a CacheWorker and
# the currently cached minions are replaced with that list
# 3. anything else is considered malformed
try:
if not new_c_data:
log.debug('ConCache Got empty update from worker')
continue
data = new_c_data[0]
if isinstance(data, six.string_types):
if data not in self.minions:
log.debug('ConCache Adding minion %s to cache',
new_c_data[0])
self.minions.append(data)
elif isinstance(data, list):
log.debug('ConCache Replacing minion list from worker')
self.minions = data
except IndexError:
log.debug('ConCache Got malformed result dict from worker')
del new_c_data
log.info('ConCache %s entries in cache', len(self.minions))
# check for next timer-event to start new jobs
if socks.get(timer_in) == zmq.POLLIN:
sec_event = serial.loads(timer_in.recv())
# update the list every 30 seconds
if int(sec_event % 30) == 0:
cw = CacheWorker(self.opts)
cw.start()
self.stop()
creq_in.close()
cupd_in.close()
timer_in.close()
context.term()
log.debug('ConCache Shutting down')
def get_id():
'''
Guess the id of the minion.
- Check /etc/hostname for a value other than localhost
- If socket.getfqdn() returns us something other than localhost, use it
- Check /etc/hosts for something that isn't localhost that maps to 127.*
- Look for a routeable / public IP
- A private IP is better than a loopback IP
- localhost may be better than killing the minion
Returns two values: the detected ID, and a boolean value noting whether or
not an IP address is being used for the ID.
'''
log.debug('Guessing ID. The id can be explicitly in set {0}'.format(
os.path.join(syspaths.CONFIG_DIR, 'minion')))
# Check /etc/hostname
try:
with salt.utils.fopen('/etc/hostname') as hfl:
name = hfl.read().strip()
if re.search(r'\s', name):
log.warning('Whitespace character detected in /etc/hostname. '
'This file should not contain any whitespace.')
else:
if name != 'localhost':
return name, False
except Exception:
pass
# Nothing in /etc/hostname or /etc/hostname not found
fqdn = socket.getfqdn()
if fqdn != 'localhost':
log.info('Found minion id from getfqdn(): {0}'.format(fqdn))
return fqdn, False
# Can /etc/hosts help us?
try:
with salt.utils.fopen('/etc/hosts') as hfl:
for line in hfl:
names = line.split()
ip_ = names.pop(0)
if ip_.startswith('127.'):
for name in names:
if name != 'localhost':
log.info(
'Found minion id in hosts file: {0}'.format(
name))
return name, False
except Exception:
pass
# Can Windows 'hosts' file help?
try:
windir = os.getenv("WINDIR")
with salt.utils.fopen(windir +
'\\system32\\drivers\\etc\\hosts') as hfl:
for line in hfl:
# skip commented or blank lines
if line[0] == '#' or len(line) <= 1:
continue
# process lines looking for '127.' in first column
try:
entry = line.split()
if entry[0].startswith('127.'):
for name in entry[1:]: # try each name in the row
if name != 'localhost':
log.info('Found minion id in hosts file: {0}'.
format(name))
return name, False
except IndexError:
pass # could not split line (malformed entry?)
except Exception:
pass
# What IP addresses do we have?
ip_addresses = [
salt.utils.network.IPv4Address(addr)
for addr in salt.utils.network.ip_addrs(include_loopback=True)
if not addr.startswith('127.')
]
for addr in ip_addresses:
if not addr.is_private:
log.info('Using public ip address for id: {0}'.format(addr))
return str(addr), True
if ip_addresses:
addr = ip_addresses.pop(0)
log.info('Using private ip address for id: {0}'.format(addr))
return str(addr), True
log.error('No id found, falling back to localhost')
return 'localhost', False
def get_elb_lbs():
"""
Returns a dictionary of load balancer names as keys
each with their respective attributes
"""
# attributes to extract from the load balancer boto objects
# this could possibly be a named argument too
extract_attrs = ['scheme', 'dns_name', 'vpc_id', 'name', 'security_groups']
try:
instance_metadata = boto.utils.get_instance_metadata(timeout=5,
num_retries=2)
except Exception as e:
log.exception("Error getting ELB names: {}".format(e))
return {'custom_grain_error': True}
# Setup the lbs grain
lbs_grain = {'lbs': {}}
# Collect details about this instance
vpc_id = instance_metadata['network']['interfaces']['macs'].values(
)[0]['vpc-id']
region = instance_metadata['placement']['availability-zone'][:-1]
# Collect load balancers of this instance (in the same vpc)
try:
elb_connection = boto.ec2.elb.connect_to_region(region)
# find load balancers by vpc_id
all_lbs = [
lb for lb in elb_connection.get_all_load_balancers()
if lb.vpc_id == vpc_id
]
log.debug(
'all lbs before filtering by instance id: {}'.format(all_lbs))
# further filter the load balancers by instance id
lbs = [
lb for lb in all_lbs for inst in lb.instances
if inst.id == instance_metadata['instance-id']
]
# initialise and populate the output of load balancers
out = {}
[out.update({l.name: {}}) for l in lbs]
[
out[l.name].update({attr: getattr(l, attr, None)})
for attr in extract_attrs for l in lbs
]
if not out:
# This loglevel could perhaps be adjusted to something more visible
log.warning("No ELBs found for this instance, this is unusual, "
"but we will not break highstate")
lbs_grain['lbs'] = out
except Exception as e:
# This prints a user-friendly error with stacktrace
log.exception("Error getting ELB names: {}".format(e))
return {'custom_grain_error': True}
return lbs_grain
