Merge pull request #20 from CIRCL/testing

Big refactoring, make the workflow much easyer to follow.
pull/21/head
Jules 2014-08-22 18:36:15 +02:00
commit 87a57e1f05
43 changed files with 724 additions and 1413 deletions

21
.gitignore vendored Normal file
View File

@ -0,0 +1,21 @@
# Temp files
*.swp
*.pyc
# Install Dirs
AILENV
redis-leveldb
redis
Blooms
LEVEL_DB_DATA
PASTES
bin/indexdir/
logs/
# Webstuff
var/www/static/
!var/www/static/css/dygraph_gallery.css
!var/www/static/js/indexjavascript.js
# Local config
bin/packages/config.cfg

View File

@ -38,6 +38,7 @@ def main():
port=cfg.getint("Redis_Queues", "port"),
db=cfg.getint("Redis_Queues", "db"))
publisher.port = 6380
publisher.channel = "Script"
create_dirfile(r_serv, args.directory, args.ow)

87
bin/Helper.py Executable file
View File

@ -0,0 +1,87 @@
#!/usr/bin/env python2
# -*-coding:UTF-8 -*
"""
Queue helper module
============================
This module subscribe to a Publisher stream and put the received messages
into a Redis-list waiting to be popped later by others scripts.
..note:: Module ZMQ_Something_Q and ZMQ_Something are closely bound, always put
the same Subscriber name in both of them.
"""
import redis
import ConfigParser
import os
import zmq
class Redis_Queues(object):
def __init__(self, conf_section, conf_channel, subscriber_name):
configfile = os.path.join(os.environ['AIL_BIN'], 'packages/config.cfg')
print configfile
if not os.path.exists(configfile):
raise Exception('Unable to find the configuration file. \
Did you set environment variables? \
Or activate the virtualenv.')
self.config = ConfigParser.ConfigParser()
self.config.read(configfile)
self.subscriber_name = subscriber_name
self.sub_channel = self.config.get(conf_section, conf_channel)
self.redis_channel = self.sub_channel + self.subscriber_name
# Redis Queue
config_section = "Redis_Queues"
self.r_queues = redis.StrictRedis(
host=self.config.get(config_section, "host"),
port=self.config.getint(config_section, "port"),
db=self.config.getint(config_section, "db"))
def zmq_sub(self, conf_section):
sub_address = self.config.get(conf_section, 'adress')
context = zmq.Context()
self.sub_socket = context.socket(zmq.SUB)
self.sub_socket.connect(sub_address)
self.sub_socket.setsockopt(zmq.SUBSCRIBE, self.sub_channel)
def zmq_pub(self, config_section, config_channel):
context = zmq.Context()
self.pub_socket = context.socket(zmq.PUB)
self.pub_socket.bind(self.config.get(config_section, 'adress'))
if config_channel is not None:
self.pub_channel = self.config.get(config_section, config_channel)
else:
# The publishing channel is defined dynamically
self.pub_channel = None
def zmq_pub_send(self, msg):
if self.pub_channel is None:
raise Exception('A channel is reqired to send a message.')
self.pub_socket.send('{} {}'.format(self.pub_channel, msg))
def redis_rpop(self):
return self.r_queues.rpop(self.redis_channel)
def redis_queue_shutdown(self, is_queue=False):
if is_queue:
flag = self.subscriber_name + '_Q'
else:
flag = self.subscriber_name
# srem returns False if the element does not exists
return self.r_queues.srem('SHUTDOWN_FLAGS', flag)
def redis_queue_subscribe(self, publisher):
publisher.info("Suscribed to channel {}".format(self.sub_channel))
while True:
msg = self.sub_socket.recv()
p = self.r_queues.pipeline()
p.sadd("queues", self.redis_channel)
p.lpush(self.redis_channel, msg)
p.execute()
if self.redis_queue_shutdown(True):
print "Shutdown Flag Up: Terminating"
publisher.warning("Shutdown Flag Up: Terminating.")
break

View File

@ -9,10 +9,13 @@ WHITE="\\033[0;02m"
YELLOW="\\033[1;33m"
CYAN="\\033[1;36m"
#Modify these PATH
export PATH=$(pwd):$PATH
export PATH=/opt/redis-2.8.12/src/:$PATH
export PATH=/opt/redis-leveldb/:$PATH
[ -z "$AIL_HOME" ] && echo "Needs the env var AIL_HOME. Run the script from the virtual environment." && exit 1;
[ -z "$AIL_REDIS" ] && echo "Needs the env var AIL_REDIS. Run the script from the virtual environment." && exit 1;
[ -z "$AIL_LEVELDB" ] && echo "Needs the env var AIL_LEVELDB. Run the script from the virtual environment." && exit 1;
export PATH=$AIL_HOME:$PATH
export PATH=$AIL_REDIS:$PATH
export PATH=$AIL_LEVELDB:$PATH
function helptext {
echo -e $YELLOW"
@ -45,7 +48,7 @@ function helptext {
}
function launching_redis {
conf_dir='/home/user/AIL-framework/configs/'
conf_dir="${AIL_HOME}/configs/"
screen -dmS "Redis"
sleep 0.1
@ -60,7 +63,7 @@ function launching_redis {
function launching_lvldb {
#Want to launch more level_db?
lvdbhost='127.0.0.1'
lvdbdir='/home/user/AIL-framework/LEVEL_DB_DATA/'
lvdbdir="${AIL_HOME}/LEVEL_DB_DATA/"
db1_y='2013'
db2_y='2014'
nb_db=13

View File

@ -33,6 +33,7 @@ def main():
db=cfg.getint("Redis_Queues", "db"))
# LOGGING #
publisher.port = 6380
publisher.channel = "Queuing"
# ZMQ #

View File

@ -34,6 +34,7 @@ def main():
db=cfg.getint("Redis_Level_DB_Hashs", "db"))
# LOGGING #
publisher.port = 6380
publisher.channel = "Graph"
# FUNCTIONS #

View File

@ -22,8 +22,9 @@ Requirements
"""
import redis
import ConfigParser
import os
configfile = './packages/config.cfg'
configfile = os.path.join(os.environ['AIL_BIN'], './packages/config.cfg')
def main():
@ -38,6 +39,7 @@ def main():
port=cfg.getint("Redis_Queues", "port"),
db=cfg.getint("Redis_Queues", "db"))
# FIXME: automatic based on the queue name.
# ### SCRIPTS ####
r_serv.sadd("SHUTDOWN_FLAGS", "Feed")
r_serv.sadd("SHUTDOWN_FLAGS", "Categ")

View File

@ -20,50 +20,35 @@ Requirements
*Need the ZMQ_Feed_Q Module running to be able to work properly.
"""
import redis
import ConfigParser
import base64
import os
import time
from pubsublogger import publisher
from packages import ZMQ_PubSub
configfile = './packages/config.cfg'
import Helper
def main():
"""Main Function"""
if __name__ == "__main__":
publisher.port = 6380
publisher.channel = "Script"
# CONFIG #
cfg = ConfigParser.ConfigParser()
cfg.read(configfile)
config_section = 'Feed'
config_channel = 'topicfilter'
subscriber_name = 'feed'
# REDIS
r_serv = redis.StrictRedis(
host=cfg.get("Redis_Queues", "host"),
port=cfg.getint("Redis_Queues", "port"),
db=cfg.getint("Redis_Queues", "db"))
h = Helper.Redis_Queues(config_section, config_channel, subscriber_name)
# ZMQ #
channel = cfg.get("Feed", "topicfilter")
# Subscriber
subscriber_name = "feed"
subscriber_config_section = "Feed"
# Publisher
publisher_name = "pubfed"
publisher_config_section = "PubSub_Global"
Sub = ZMQ_PubSub.ZMQSub(configfile, subscriber_config_section, channel, subscriber_name)
PubGlob = ZMQ_PubSub.ZMQPub(configfile, publisher_config_section, publisher_name)
pub_config_section = "PubSub_Global"
pub_config_channel = 'channel'
h.zmq_pub(pub_config_section, pub_config_channel)
# LOGGING #
publisher.channel = "Script"
publisher.info("Feed Script started to receive & publish.")
while True:
message = Sub.get_msg_from_queue(r_serv)
message = h.redis_rpop()
# Recovering the streamed message informations.
if message is not None:
if len(message.split()) == 3:
@ -75,8 +60,7 @@ def main():
publisher.debug("Empty Paste: {0} not processed".format(paste))
continue
else:
if r_serv.sismember("SHUTDOWN_FLAGS", "Feed"):
r_serv.srem("SHUTDOWN_FLAGS", "Feed")
if h.redis_queue_shutdown():
print "Shutdown Flag Up: Terminating"
publisher.warning("Shutdown Flag Up: Terminating.")
break
@ -84,24 +68,13 @@ def main():
time.sleep(10)
continue
# Creating the full filepath
filename = cfg.get("Directories", "pastes") + paste
filename = os.path.join(os.environ['AIL_HOME'],
h.config.get("Directories", "pastes"), paste)
dirname = os.path.dirname(filename)
if not os.path.exists(dirname):
os.makedirs(dirname)
if not os.path.exists(filename.rsplit("/", 1)[0]):
os.makedirs(filename.rsplit("/", 1)[0])
else:
# Path already existing
pass
with open(filename, 'wb') as f:
f.write(base64.standard_b64decode(gzip64encoded))
decoded_gzip = base64.standard_b64decode(gzip64encoded)
# paste, zlib.decompress(decoded_gzip, zlib.MAX_WBITS|16)
with open(filename, 'wb') as F:
F.write(decoded_gzip)
msg = cfg.get("PubSub_Global", "channel")+" "+filename
PubGlob.send_message(msg)
publisher.debug("{0} Published".format(msg))
if __name__ == "__main__":
main()
h.zmq_pub_send(filename)

View File

@ -20,45 +20,19 @@ Requirements
"channel_name"+" "+/path/to/the/paste.gz+" "base64_data_encoded_paste"
"""
import redis
import ConfigParser
from pubsublogger import publisher
from packages import ZMQ_PubSub
configfile = './packages/config.cfg'
import Helper
def main():
"""Main Function"""
# CONFIG #
cfg = ConfigParser.ConfigParser()
cfg.read(configfile)
# REDIS #
r_serv = redis.StrictRedis(
host=cfg.get("Redis_Queues", "host"),
port=cfg.getint("Redis_Queues", "port"),
db=cfg.getint("Redis_Queues", "db"))
# LOGGING #
publisher.channel = "Queuing"
# ZMQ #
channel = cfg.get("Feed", "topicfilter")
sub = ZMQ_PubSub.ZMQSub(configfile, "Feed", channel, "feed")
# FUNCTIONS #
publisher.info("""Suscribed to channel {0}""".format(channel))
while True:
sub.get_and_lpush(r_serv)
if r_serv.sismember("SHUTDOWN_FLAGS", "Feed_Q"):
r_serv.srem("SHUTDOWN_FLAGS", "Feed_Q")
print "Shutdown Flag Up: Terminating"
publisher.warning("Shutdown Flag Up: Terminating.")
break
if __name__ == "__main__":
main()
publisher.channel = "Queuing"
config_section = 'Feed'
config_channel = 'topicfilter'
subscriber_name = 'feed'
h = Helper.Redis_Queues(config_section, config_channel, subscriber_name)
h.zmq_sub(config_section)
h.redis_queue_subscribe(publisher)

View File

@ -36,111 +36,77 @@ Requirements
*Need the ZMQ_PubSub_Tokenize_Q Module running to be able to work properly.
"""
import redis
import os
import argparse
import ConfigParser
import time
from packages import ZMQ_PubSub
from pubsublogger import publisher
from packages import Paste
configfile = './packages/config.cfg'
import Helper
if __name__ == "__main__":
publisher.port = 6380
publisher.channel = "Script"
def main():
"""Main Function"""
config_section = 'PubSub_Words'
config_channel = 'channel_0'
subscriber_name = 'categ'
# CONFIG #
cfg = ConfigParser.ConfigParser()
cfg.read(configfile)
h = Helper.Redis_Queues(config_section, config_channel, subscriber_name)
# Publisher
pub_config_section = 'PubSub_Categ'
h.zmq_pub(pub_config_section, None)
# SCRIPT PARSER #
parser = argparse.ArgumentParser(
description='''This script is a part of the Analysis Information Leak framework.''',
epilog='''''')
description='This script is a part of the Analysis Information \
Leak framework.')
parser.add_argument(
'-l', type=str, default="../files/list_categ_files",
help='Path to the list_categ_files (../files/list_categ_files)',
'-d', type=str, default="../files/",
help='Path to the directory containing the category files.',
action='store')
args = parser.parse_args()
# REDIS #
r_serv = redis.StrictRedis(
host=cfg.get("Redis_Queues", "host"),
port=cfg.getint("Redis_Queues", "port"),
db=cfg.getint("Redis_Queues", "db"))
# LOGGING #
publisher.channel = "Script"
# ZMQ #
channel = cfg.get("PubSub_Words", "channel_0")
subscriber_name = "categ"
subscriber_config_section = "PubSub_Words"
publisher_name = "pubcateg"
publisher_config_section = "PubSub_Categ"
sub = ZMQ_PubSub.ZMQSub(configfile, subscriber_config_section, channel,
subscriber_name)
pub = ZMQ_PubSub.ZMQPub(configfile, publisher_config_section,
publisher_name)
# FUNCTIONS #
publisher.info("Script Categ subscribed to channel {0}".format(
cfg.get("PubSub_Words", "channel_0")))
publisher.info(
"Script Categ subscribed to channel {}".format(h.sub_channel))
with open(args.l, 'rb') as L:
tmp_dict = {}
tmp_dict = {}
for filename in os.listdir(args.d):
bname = os.path.basename(filename)
tmp_dict[bname] = []
with open(os.path.join(args.d, filename), 'r') as f:
for l in f:
tmp_dict[bname].append(l.strip())
for num, fname in enumerate(L):
# keywords temp list
tmp_list = []
with open(fname[:-1], 'rb') as LS:
for num, kword in enumerate(LS):
tmp_list.append(kword[:-1])
tmp_dict[fname.split('/')[-1][:-1]] = tmp_list
message = sub.get_msg_from_queue(r_serv)
prec_filename = None
while True:
message = h.redis_rpop()
if message is not None:
channel, filename, word, score = message.split()
if prec_filename is None or filename != prec_filename:
PST = Paste.Paste(filename)
prec_filename = filename
prec_filename = filename
for categ, words_list in tmp_dict.items():
for categ, list in tmp_dict.items():
if word.lower() in list:
channel = categ
msg = channel+" "+PST.p_path+" "+word+" "+score
pub.send_message(msg)
# dico_categ.add(categ)
if word.lower() in words_list:
h.pub_channel = categ
h.zmq_pub_send('{} {} {}'.format(PST.p_path, word, score))
publisher.info(
'Categ;{};{};{};Detected {} "{}"'.format(
PST.p_source, PST.p_date, PST.p_name, score, word))
else:
if r_serv.sismember("SHUTDOWN_FLAGS", "Categ"):
r_serv.srem("SHUTDOWN_FLAGS", "Categ")
if h.redis_queue_shutdown():
print "Shutdown Flag Up: Terminating"
publisher.warning("Shutdown Flag Up: Terminating.")
break
publisher.debug("Script Categ is Idling 10s")
time.sleep(10)
message = sub.get_msg_from_queue(r_serv)
if __name__ == "__main__":
main()

View File

@ -17,47 +17,20 @@ Requirements
*Should register to the Publisher "ZMQ_PubSub_Tokenize"
"""
import redis
import ConfigParser
from pubsublogger import publisher
from packages import ZMQ_PubSub
configfile = './packages/config.cfg'
import Helper
def main():
"""Main Function"""
# CONFIG #
cfg = ConfigParser.ConfigParser()
cfg.read(configfile)
# REDIS #
r_serv = redis.StrictRedis(
host=cfg.get("Redis_Queues", "host"),
port=cfg.getint("Redis_Queues", "port"),
db=cfg.getint("Redis_Queues", "db"))
# LOGGING #
publisher.channel = "Queuing"
# ZMQ #
channel = cfg.get("PubSub_Words", "channel_0")
subscriber_name = "categ"
subscriber_config_section = "PubSub_Words"
sub = ZMQ_PubSub.ZMQSub(configfile, subscriber_config_section, channel, subscriber_name)
# FUNCTIONS #
publisher.info("""Suscribed to channel {0}""".format(channel))
while True:
sub.get_and_lpush(r_serv)
if r_serv.sismember("SHUTDOWN_FLAGS", "Categ_Q"):
r_serv.srem("SHUTDOWN_FLAGS", "Categ_Q")
print "Shutdown Flag Up: Terminating"
publisher.warning("Shutdown Flag Up: Terminating.")
break
if __name__ == "__main__":
main()
publisher.port = 6380
publisher.channel = 'Queuing'
config_section = 'PubSub_Words'
config_channel = 'channel_0'
subscriber_name = 'categ'
h = Helper.Redis_Queues(config_section, config_channel, subscriber_name)
h.zmq_sub(config_section)
h.redis_queue_subscribe(publisher)

View File

@ -5,10 +5,11 @@
The ZMQ_PubSub_Lines Module
============================
This module is consuming the Redis-list created by the ZMQ_PubSub_Line_Q Module.
This module is consuming the Redis-list created by the ZMQ_PubSub_Line_Q
Module.
It perform a sorting on the line's length and publish/forward them to differents
channels:
It perform a sorting on the line's length and publish/forward them to
differents channels:
*Channel 1 if max length(line) < max
*Channel 2 if max length(line) > max
@ -26,81 +27,57 @@ Requirements
*Need the ZMQ_PubSub_Line_Q Module running to be able to work properly.
"""
import redis
import argparse
import ConfigParser
import time
from packages import Paste
from packages import ZMQ_PubSub
from pubsublogger import publisher
configfile = './packages/config.cfg'
import Helper
if __name__ == "__main__":
publisher.port = 6380
publisher.channel = "Script"
def main():
"""Main Function"""
config_section = 'PubSub_Global'
config_channel = 'channel'
subscriber_name = 'line'
# CONFIG #
cfg = ConfigParser.ConfigParser()
cfg.read(configfile)
h = Helper.Redis_Queues(config_section, config_channel, subscriber_name)
# Publisher
pub_config_section = 'PubSub_Longlines'
h.zmq_pub(pub_config_section, None)
# Subscriber
h.zmq_sub(config_section)
# SCRIPT PARSER #
parser = argparse.ArgumentParser(
description='''This script is a part of the Analysis Information Leak framework.''',
epilog='''''')
description='''This script is a part of the Analysis Information \
Leak framework.''')
parser.add_argument('-max', type=int, default=500,
help='The limit between "short lines" and "long lines" (500)',
action='store')
parser.add_argument(
'-max', type=int, default=500,
help='The limit between "short lines" and "long lines"',
action='store')
args = parser.parse_args()
# REDIS #
r_serv = redis.StrictRedis(
host=cfg.get("Redis_Data_Merging", "host"),
port=cfg.getint("Redis_Data_Merging", "port"),
db=cfg.getint("Redis_Data_Merging", "db"))
r_serv1 = redis.StrictRedis(
host=cfg.get("Redis_Queues", "host"),
port=cfg.getint("Redis_Queues", "port"),
db=cfg.getint("Redis_Queues", "db"))
# LOGGING #
publisher.channel = "Script"
# ZMQ #
# Subscriber
channel = cfg.get("PubSub_Global", "channel")
subscriber_name = "line"
subscriber_config_section = "PubSub_Global"
# Publisher
publisher_config_section = "PubSub_Longlines"
publisher_name = "publine"
sub = ZMQ_PubSub.ZMQSub(configfile, subscriber_config_section, channel, subscriber_name)
pub = ZMQ_PubSub.ZMQPub(configfile, publisher_config_section, publisher_name)
channel_0 = cfg.get("PubSub_Longlines", "channel_0")
channel_1 = cfg.get("PubSub_Longlines", "channel_1")
channel_0 = h.config.get("PubSub_Longlines", "channel_0")
channel_1 = h.config.get("PubSub_Longlines", "channel_1")
# FUNCTIONS #
tmp_string = "Lines script Subscribed to channel {} and Start to publish on channel {}, {}"
publisher.info(tmp_string.format(
cfg.get("PubSub_Global", "channel"),
cfg.get("PubSub_Longlines", "channel_0"),
cfg.get("PubSub_Longlines", "channel_1")))
tmp_string = "Lines script Subscribed to channel {} and Start to publish \
on channel {}, {}"
publisher.info(tmp_string.format(h.sub_channel, channel_0, channel_1))
while True:
try:
message = sub.get_msg_from_queue(r_serv1)
message = h.redis_rpop()
if message is not None:
PST = Paste.Paste(message.split(" ", -1)[-1])
else:
if r_serv1.sismember("SHUTDOWN_FLAGS", "Lines"):
r_serv1.srem("SHUTDOWN_FLAGS", "Lines")
if h.redis_queue_shutdown():
print "Shutdown Flag Up: Terminating"
publisher.warning("Shutdown Flag Up: Terminating.")
break
@ -108,21 +85,17 @@ def main():
time.sleep(10)
continue
# FIXME do it in the paste class
lines_infos = PST.get_lines_info()
PST.save_attribute_redis("p_nb_lines", lines_infos[0])
PST.save_attribute_redis("p_max_length_line", lines_infos[1])
PST.save_attribute_redis(r_serv, "p_nb_lines", lines_infos[0])
PST.save_attribute_redis(r_serv, "p_max_length_line", lines_infos[1])
r_serv.sadd("Pastes_Objects", PST.p_path)
# FIXME Not used.
PST.store.sadd("Pastes_Objects", PST.p_path)
if lines_infos[1] >= args.max:
msg = channel_0+" "+PST.p_path
h.pub_channel = channel_0
else:
msg = channel_1+" "+PST.p_path
pub.send_message(msg)
h.pub_channel = channel_1
h.zmq_pub_send(PST.p_path)
except IOError:
print "CRC Checksum Error on : ", PST.p_path
pass
if __name__ == "__main__":
main()

View File

@ -18,47 +18,18 @@ Requirements
"""
import redis
import ConfigParser
from pubsublogger import publisher
from packages import ZMQ_PubSub
configfile = './packages/config.cfg'
def main():
"""Main Function"""
# CONFIG #
cfg = ConfigParser.ConfigParser()
cfg.read(configfile)
# REDIS #
r_serv = redis.StrictRedis(
host=cfg.get("Redis_Queues", "host"),
port=cfg.getint("Redis_Queues", "port"),
db=cfg.getint("Redis_Queues", "db"))
# LOGGING #
publisher.channel = "Queuing"
# ZMQ #
channel = cfg.get("PubSub_Global", "channel")
subscriber_name = "line"
sub = ZMQ_PubSub.ZMQSub(configfile, "PubSub_Global", channel, subscriber_name)
# FUNCTIONS #
publisher.info("""Suscribed to channel {0}""".format(channel))
while True:
sub.get_and_lpush(r_serv)
if r_serv.sismember("SHUTDOWN_FLAGS", "Lines_Q"):
r_serv.srem("SHUTDOWN_FLAGS", "Lines_Q")
print "Shutdown Flag Up: Terminating"
publisher.warning("Shutdown Flag Up: Terminating.")
break
import Helper
if __name__ == "__main__":
main()
publisher.port = 6380
publisher.channel = "Queuing"
config_section = "PubSub_Global"
config_channel = 'channel'
subscriber_name = 'line'
h = Helper.Redis_Queues(config_section, config_channel, subscriber_name)
h.zmq_sub(config_section)
h.redis_queue_subscribe(publisher)

View File

@ -4,9 +4,11 @@
The ZMQ_PubSub_Lines Module
============================
This module is consuming the Redis-list created by the ZMQ_PubSub_Tokenize_Q Module.
This module is consuming the Redis-list created by the ZMQ_PubSub_Tokenize_Q
Module.
It tokenize the content of the paste and publish the result in the following format:
It tokenize the content of the paste and publish the result in the following
format:
channel_name+' '+/path/of/the/paste.gz+' '+tokenized_word+' '+scoring
..seealso:: Paste method (_get_top_words)
@ -21,72 +23,44 @@ Requirements
*Need the ZMQ_PubSub_Tokenize_Q Module running to be able to work properly.
"""
import redis
import ConfigParser
import time
from packages import Paste
from packages import ZMQ_PubSub
from pubsublogger import publisher
configfile = './packages/config.cfg'
import Helper
def main():
"""Main Function"""
# CONFIG #
cfg = ConfigParser.ConfigParser()
cfg.read(configfile)
# REDIS #
r_serv = redis.StrictRedis(
host=cfg.get("Redis_Queues", "host"),
port=cfg.getint("Redis_Queues", "port"),
db=cfg.getint("Redis_Queues", "db"))
# LOGGING #
if __name__ == "__main__":
publisher.port = 6380
publisher.channel = "Script"
# ZMQ #
channel = cfg.get("PubSub_Longlines", "channel_1")
subscriber_name = "tokenize"
subscriber_config_section = "PubSub_Longlines"
config_section = 'PubSub_Longlines'
config_channel = 'channel_1'
subscriber_name = 'tokenize'
h = Helper.Redis_Queues(config_section, config_channel, subscriber_name)
# Publisher
publisher_config_section = "PubSub_Words"
publisher_name = "pubtokenize"
pub_config_section = 'PubSub_Words'
pub_config_channel = 'channel_0'
h.zmq_pub(pub_config_section, pub_config_channel)
sub = ZMQ_PubSub.ZMQSub(configfile, subscriber_config_section, channel, subscriber_name)
pub = ZMQ_PubSub.ZMQPub(configfile, publisher_config_section, publisher_name)
channel_0 = cfg.get("PubSub_Words", "channel_0")
# FUNCTIONS #
publisher.info("Tokeniser subscribed to channel {0}".format(cfg.get("PubSub_Longlines", "channel_1")))
# LOGGING #
publisher.info("Tokeniser subscribed to channel {}".format(h.sub_channel))
while True:
message = sub.get_msg_from_queue(r_serv)
message = h.redis_rpop()
print message
if message is not None:
PST = Paste.Paste(message.split(" ", -1)[-1])
paste = Paste.Paste(message.split(" ", -1)[-1])
for word, score in paste._get_top_words().items():
if len(word) >= 4:
h.zmq_pub_send('{} {} {}'.format(paste.p_path, word,
score))
else:
if r_serv.sismember("SHUTDOWN_FLAGS", "Tokenize"):
r_serv.srem("SHUTDOWN_FLAGS", "Tokenize")
if h.redis_queue_shutdown():
print "Shutdown Flag Up: Terminating"
publisher.warning("Shutdown Flag Up: Terminating.")
break
publisher.debug("Tokeniser is idling 10s")
time.sleep(10)
print "sleepin"
continue
for word, score in PST._get_top_words().items():
if len(word) >= 4:
msg = channel_0+' '+PST.p_path+' '+str(word)+' '+str(score)
pub.send_message(msg)
print msg
else:
pass
if __name__ == "__main__":
main()

View File

@ -17,48 +17,20 @@ Requirements
*Should register to the Publisher "ZMQ_PubSub_Line" channel 1
"""
import redis
import ConfigParser
from pubsublogger import publisher
from packages import ZMQ_PubSub
configfile = './packages/config.cfg'
import Helper
def main():
"""Main Function"""
# CONFIG #
cfg = ConfigParser.ConfigParser()
cfg.read(configfile)
# REDIS #
r_serv = redis.StrictRedis(
host=cfg.get("Redis_Queues", "host"),
port=cfg.getint("Redis_Queues", "port"),
db=cfg.getint("Redis_Queues", "db"))
# LOGGING #
publisher.channel = "Queuing"
# ZMQ #
channel = cfg.get("PubSub_Longlines", "channel_1")
subscriber_name = "tokenize"
subscriber_config_section = "PubSub_Longlines"
sub = ZMQ_PubSub.ZMQSub(configfile, subscriber_config_section, channel, subscriber_name)
# FUNCTIONS #
publisher.info("""Suscribed to channel {0}""".format(channel))
while True:
sub.get_and_lpush(r_serv)
if r_serv.sismember("SHUTDOWN_FLAGS", "Tokenize_Q"):
r_serv.srem("SHUTDOWN_FLAGS", "Tokenize_Q")
print "Shutdown Flag Up: Terminating"
publisher.warning("Shutdown Flag Up: Terminating.")
break
if __name__ == "__main__":
main()
publisher.port = 6380
publisher.channel = "Queuing"
config_section = 'PubSub_Longlines'
config_channel = 'channel_1'
subscriber_name = 'tokenize'
h = Helper.Redis_Queues(config_section, config_channel, subscriber_name)
h.zmq_sub(config_section)
h.redis_queue_subscribe(publisher)

View File

@ -21,49 +21,34 @@ Requirements
*Need running Redis instances. (Redis)
"""
import redis
import ConfigParser
import time
from packages import ZMQ_PubSub
from pubsublogger import publisher
configfile = './packages/config.cfg'
import Helper
def main():
"""Main Function"""
# CONFIG #
cfg = ConfigParser.ConfigParser()
cfg.read('./packages/config.cfg')
# REDIS #
r_serv = redis.StrictRedis(
host=cfg.get("Redis_Queues", "host"),
port=cfg.getint("Redis_Queues", "port"),
db=cfg.getint("Redis_Queues", "db"))
# LOGGING #
if __name__ == "__main__":
publisher.port = 6380
publisher.channel = "Global"
# ZMQ #
pub_glob = ZMQ_PubSub.ZMQPub(configfile, "PubSub_Global", "global")
config_section = 'PubSub_Global'
config_channel = 'channel'
subscriber_name = 'global'
# FONCTIONS #
h = Helper.Redis_Queues(config_section, config_channel, subscriber_name)
# Publisher
pub_config_section = 'PubSub_Global'
pub_config_channel = 'channel'
h.zmq_pub(pub_config_section, pub_config_channel)
# LOGGING #
publisher.info("Starting to publish.")
while True:
filename = r_serv.lpop("filelist")
filename = h.redis_rpop()
if filename is not None:
msg = cfg.get("PubSub_Global", "channel")+" "+filename
pub_glob.send_message(msg)
publisher.debug("{0} Published".format(msg))
h.zmq_pub_send(filename)
else:
time.sleep(10)
publisher.debug("Nothing to publish")
if __name__ == "__main__":
main()

View File

@ -5,10 +5,10 @@
The ZMQ_Sub_Attribute Module
============================
This module is consuming the Redis-list created by the ZMQ_PubSub_Line_Q Module.
This module is consuming the Redis-list created by the ZMQ_PubSub_Line_Q Module
It perform a sorting on the line's length and publish/forward them to differents
channels:
It perform a sorting on the line's length and publish/forward them to
differents channels:
*Channel 1 if max length(line) < max
*Channel 2 if max length(line) > max
@ -26,57 +26,36 @@ Requirements
*Need the ZMQ_PubSub_Line_Q Module running to be able to work properly.
"""
import redis
import ConfigParser
import time
from packages import Paste
from packages import ZMQ_PubSub
from pubsublogger import publisher
configfile = './packages/config.cfg'
import Helper
def main():
"""Main Function"""
# CONFIG #
cfg = ConfigParser.ConfigParser()
cfg.read(configfile)
# REDIS #
r_serv = redis.StrictRedis(
host=cfg.get("Redis_Data_Merging", "host"),
port=cfg.getint("Redis_Data_Merging", "port"),
db=cfg.getint("Redis_Data_Merging", "db"))
r_serv1 = redis.StrictRedis(
host=cfg.get("Redis_Queues", "host"),
port=cfg.getint("Redis_Queues", "port"),
db=cfg.getint("Redis_Queues", "db"))
# LOGGING #
if __name__ == "__main__":
publisher.port = 6380
publisher.channel = "Script"
# ZMQ #
# Subscriber
channel = cfg.get("PubSub_Global", "channel")
subscriber_name = "attributes"
subscriber_config_section = "PubSub_Global"
config_section = 'PubSub_Global'
config_channel = 'channel'
subscriber_name = 'attributes'
sub = ZMQ_PubSub.ZMQSub(configfile, subscriber_config_section, channel, subscriber_name)
h = Helper.Redis_Queues(config_section, config_channel, subscriber_name)
# Subscriber
h.zmq_sub(config_section)
# FUNCTIONS #
publisher.info("""ZMQ Attribute is Running""")
while True:
try:
message = sub.get_msg_from_queue(r_serv1)
message = h.redis_rpop()
if message is not None:
PST = Paste.Paste(message.split(" ", -1)[-1])
else:
if r_serv1.sismember("SHUTDOWN_FLAGS", "Attributes"):
r_serv1.srem("SHUTDOWN_FLAGS", "Attributes")
if h.redis_queue_shutdown():
print "Shutdown Flag Up: Terminating"
publisher.warning("Shutdown Flag Up: Terminating.")
break
@ -84,20 +63,14 @@ def main():
time.sleep(10)
continue
encoding = PST._get_p_encoding()
language = PST._get_p_language()
PST.save_attribute_redis(r_serv, "p_encoding", encoding)
PST.save_attribute_redis(r_serv, "p_language", language)
r_serv.sadd("Pastes_Objects", PST.p_path)
PST.save_all_attributes_redis(r_serv)
# FIXME do it directly in the class
PST.save_attribute_redis("p_encoding", PST._get_p_encoding())
PST.save_attribute_redis("p_language", PST._get_p_language())
# FIXME why not all saving everything there.
PST.save_all_attributes_redis()
# FIXME Not used.
PST.store.sadd("Pastes_Objects", PST.p_path)
except IOError:
print "CRC Checksum Failed on :", PST.p_path
publisher.error('{0};{1};{2};{3};{4}'.format("Duplicate", PST.p_source, PST.p_date, PST.p_name, "CRC Checksum Failed"))
pass
if __name__ == "__main__":
main()
publisher.error('Duplicate;{};{};{};CRC Checksum Failed'.format(
PST.p_source, PST.p_date, PST.p_name))

View File

@ -18,47 +18,19 @@ Requirements
"""
import redis
import ConfigParser
from pubsublogger import publisher
from packages import ZMQ_PubSub
configfile = './packages/config.cfg'
import Helper
def main():
"""Main Function"""
# CONFIG #
cfg = ConfigParser.ConfigParser()
cfg.read(configfile)
# REDIS #
r_serv = redis.StrictRedis(
host=cfg.get("Redis_Queues", "host"),
port=cfg.getint("Redis_Queues", "port"),
db=cfg.getint("Redis_Queues", "db"))
# LOGGING #
publisher.channel = "Queuing"
# ZMQ #
channel = cfg.get("PubSub_Global", "channel")
subscriber_name = "attributes"
sub = ZMQ_PubSub.ZMQSub(configfile, "PubSub_Global", channel, subscriber_name)
# FUNCTIONS #
publisher.info("""Suscribed to channel {0}""".format(channel))
while True:
sub.get_and_lpush(r_serv)
if r_serv.sismember("SHUTDOWN_FLAGS", "Attributes_Q"):
r_serv.srem("SHUTDOWN_FLAGS", "Attributes_Q")
print "Shutdown Flag Up: Terminating"
publisher.warning("Shutdown Flag Up: Terminating.")
break
if __name__ == "__main__":
main()
publisher.port = 6380
publisher.channel = "Queuing"
config_section = 'PubSub_Global'
config_channel = 'channel'
subscriber_name = 'attributes'
h = Helper.Redis_Queues(config_section, config_channel, subscriber_name)
h.zmq_sub(config_section)
h.redis_queue_subscribe(publisher)

View File

@ -1,53 +1,38 @@
#!/usr/bin/env python2
# -*-coding:UTF-8 -*
import redis
import ConfigParser
import pprint
import time
from packages import Paste
from packages import lib_refine
from packages import ZMQ_PubSub
from pubsublogger import publisher
import Helper
configfile = './packages/config.cfg'
def main():
"""Main Function"""
# CONFIG #
cfg = ConfigParser.ConfigParser()
cfg.read(configfile)
# REDIS #
r_serv = redis.StrictRedis(
host=cfg.get("Redis_Queues", "host"),
port=cfg.getint("Redis_Queues", "port"),
db=cfg.getint("Redis_Queues", "db"))
r_serv1 = redis.StrictRedis(
host=cfg.get("Redis_Data_Merging", "host"),
port=cfg.getint("Redis_Data_Merging", "port"),
db=cfg.getint("Redis_Data_Merging", "db"))
# LOGGING #
if __name__ == "__main__":
publisher.port = 6380
publisher.channel = "Script"
# ZMQ #
sub = ZMQ_PubSub.ZMQSub(configfile, "PubSub_Categ", "creditcard_categ", "cards")
config_section = 'PubSub_Categ'
config_channel = 'channel_0'
subscriber_name = 'cards'
h = Helper.Redis_Queues(config_section, config_channel, subscriber_name)
# Subscriber
h.zmq_sub(config_section)
# FUNCTIONS #
publisher.info("Creditcard script subscribed to channel creditcard_categ")
message = sub.get_msg_from_queue(r_serv)
message = h.redis_rpop()
prec_filename = None
creditcard_regex = "4[0-9]{12}(?:[0-9]{3})?"
# mastercard_regex = "5[1-5]\d{2}([\ \-]?)\d{4}\1\d{4}\1\d{4}"
# visa_regex = "4\d{3}([\ \-]?)\d{4}\1\d{4}\1\d{4}"
# discover_regex = "6(?:011\d\d|5\d{4}|4[4-9]\d{3}|22(?:1(?:2[6-9]|[3-9]\d)|[2-8]\d\d|9(?:[01]\d|2[0-5])))\d{10}"
# discover_regex = "6(?:011\d\d|5\d{4}|4[4-9]\d{3}|22(?:1(?:2[6-9]|
# [3-9]\d)|[2-8]\d\d|9(?:[01]\d|2[0-5])))\d{10}"
# jcb_regex = "35(?:2[89]|[3-8]\d)([\ \-]?)\d{4}\1\d{4}\1\d{4}"
# amex_regex = "3[47]\d\d([\ \-]?)\d{6}\1\d{5}"
# chinaUP_regex = "62[0-5]\d{13,16}"
@ -66,28 +51,25 @@ def main():
creditcard_set.add(x)
PST.__setattr__(channel, creditcard_set)
PST.save_attribute_redis(r_serv1, channel, creditcard_set)
PST.save_attribute_redis(channel, creditcard_set)
pprint.pprint(creditcard_set)
to_print = 'CreditCard;{};{};{};'.format(PST.p_source, PST.p_date, PST.p_name)
to_print = 'CreditCard;{};{};{};'.format(
PST.p_source, PST.p_date, PST.p_name)
if (len(creditcard_set) > 0):
publisher.critical('{}Checked {} valid number(s)'.format(to_print, len(creditcard_set)))
publisher.critical('{}Checked {} valid number(s)'.format(
to_print, len(creditcard_set)))
else:
publisher.info('{}CreditCard related'.format(to_print))
prec_filename = filename
else:
if r_serv.sismember("SHUTDOWN_FLAGS", "Creditcards"):
r_serv.srem("SHUTDOWN_FLAGS", "Creditcards")
if h.redis_queue_shutdown():
print "Shutdown Flag Up: Terminating"
publisher.warning("Shutdown Flag Up: Terminating.")
break
publisher.debug("Script creditcard is idling 1m")
time.sleep(60)
message = sub.get_msg_from_queue(r_serv)
if __name__ == "__main__":
main()
message = h.redis_rpop()

View File

@ -1,44 +1,19 @@
#!/usr/bin/env python2
# -*-coding:UTF-8 -*
import redis
import ConfigParser
from packages import ZMQ_PubSub
from pubsublogger import publisher
configfile = './packages/config.cfg'
import Helper
def main():
"""Main Function"""
# CONFIG #
cfg = ConfigParser.ConfigParser()
cfg.read(configfile)
# REDIS #
r_serv = redis.StrictRedis(
host=cfg.get("Redis_Queues", "host"),
port=cfg.getint("Redis_Queues", "port"),
db=cfg.getint("Redis_Queues", "db"))
# LOGGING #
publisher.channel = "Queuing"
# ZMQ #
sub = ZMQ_PubSub.ZMQSub(configfile, "PubSub_Categ", "creditcard_categ", "cards")
# FUNCTIONS #
publisher.info("""Suscribed to channel {0}""".format("creditcard_categ"))
while True:
sub.get_and_lpush(r_serv)
if r_serv.sismember("SHUTDOWN_FLAGS", "Creditcards_Q"):
r_serv.srem("SHUTDOWN_FLAGS", "Creditcards_Q")
print "Shutdown Flag Up: Terminating"
publisher.warning("Shutdown Flag Up: Terminating.")
break
if __name__ == "__main__":
main()
publisher.port = 6380
publisher.channel = "Queuing"
config_section = 'PubSub_Categ'
config_channel = 'channel_0'
subscriber_name = 'cards'
h = Helper.Redis_Queues(config_section, config_channel, subscriber_name)
h.zmq_sub(config_section)
h.redis_queue_subscribe(publisher)

View File

@ -6,7 +6,8 @@ The ZMQ_Sub_Curve Module
This module is consuming the Redis-list created by the ZMQ_Sub_Curve_Q Module.
This modules update a .csv file used to draw curves representing selected words and their occurency per day.
This modules update a .csv file used to draw curves representing selected
words and their occurency per day.
..note:: The channel will have the name of the file created.
@ -22,72 +23,64 @@ Requirements
"""
import redis
import ConfigParser
import time
from packages import Paste as P
from packages import ZMQ_PubSub
from packages import Paste
from pubsublogger import publisher
from packages import lib_words
import os
configfile = './packages/config.cfg'
import Helper
def main():
"""Main Function"""
# CONFIG #
cfg = ConfigParser.ConfigParser()
cfg.read(configfile)
# REDIS #
r_serv = redis.StrictRedis(
host=cfg.get("Redis_Queues", "host"),
port=cfg.getint("Redis_Queues", "port"),
db=cfg.getint("Redis_Queues", "db"))
r_serv1 = redis.StrictRedis(
host=cfg.get("Redis_Level_DB", "host"),
port=cfg.get("Redis_Level_DB", "port"),
db=0)
# LOGGING #
if __name__ == "__main__":
publisher.port = 6380
publisher.channel = "Script"
# ZMQ #
channel = cfg.get("PubSub_Words", "channel_0")
config_section = 'PubSub_Words'
config_channel = 'channel_0'
subscriber_name = "curve"
subscriber_config_section = "PubSub_Words"
sub = ZMQ_PubSub.ZMQSub(configfile, subscriber_config_section, channel, subscriber_name)
h = Helper.Redis_Queues(config_section, config_channel, subscriber_name)
# Subscriber
h.zmq_sub(config_section)
# REDIS #
r_serv1 = redis.StrictRedis(
host=h.config.get("Redis_Level_DB", "host"),
port=h.config.get("Redis_Level_DB", "port"),
db=h.config.get("Redis_Level_DB", "db"))
# FUNCTIONS #
publisher.info("Script Curve subscribed to channel {0}".format(cfg.get("PubSub_Words", "channel_0")))
publisher.info("Script Curve subscribed to {}".format(h.sub_channel))
# FILE CURVE SECTION #
csv_path = cfg.get("Directories", "wordtrending_csv")
wordfile_path = cfg.get("Directories", "wordsfile")
csv_path = os.path.join(os.environ['AIL_HOME'],
h.config.get("Directories", "wordtrending_csv"))
wordfile_path = os.path.join(os.environ['AIL_HOME'],
h.config.get("Directories", "wordsfile"))
message = sub.get_msg_from_queue(r_serv)
message = h.redis_rpop()
prec_filename = None
while True:
if message is not None:
channel, filename, word, score = message.split()
if prec_filename is None or filename != prec_filename:
PST = P.Paste(filename)
lib_words.create_curve_with_word_file(r_serv1, csv_path, wordfile_path, int(PST.p_date.year), int(PST.p_date.month))
PST = Paste.Paste(filename)
lib_words.create_curve_with_word_file(
r_serv1, csv_path, wordfile_path, int(PST.p_date.year),
int(PST.p_date.month))
prec_filename = filename
prev_score = r_serv1.hget(word.lower(), PST.p_date)
print prev_score
if prev_score is not None:
r_serv1.hset(word.lower(), PST.p_date, int(prev_score) + int(score))
r_serv1.hset(word.lower(), PST.p_date,
int(prev_score) + int(score))
else:
r_serv1.hset(word.lower(), PST.p_date, score)
# r_serv.expire(word,86400) #1day
else:
if r_serv.sismember("SHUTDOWN_FLAGS", "Curve"):
r_serv.srem("SHUTDOWN_FLAGS", "Curve")
if h.redis_queue_shutdown():
print "Shutdown Flag Up: Terminating"
publisher.warning("Shutdown Flag Up: Terminating.")
break
@ -95,8 +88,4 @@ def main():
print "sleepin"
time.sleep(1)
message = sub.get_msg_from_queue(r_serv)
if __name__ == "__main__":
main()
message = h.redis_rpop()

View File

@ -17,47 +17,20 @@ Requirements
*Should register to the Publisher "ZMQ_PubSub_Tokenize"
"""
import redis
import ConfigParser
from pubsublogger import publisher
from packages import ZMQ_PubSub
configfile = './packages/config.cfg'
import Helper
def main():
"""Main Function"""
# CONFIG #
cfg = ConfigParser.ConfigParser()
cfg.read(configfile)
# REDIS #
r_serv = redis.StrictRedis(
host=cfg.get("Redis_Queues", "host"),
port=cfg.getint("Redis_Queues", "port"),
db=cfg.getint("Redis_Queues", "db"))
# LOGGING #
publisher.channel = "Queuing"
# ZMQ #
channel = cfg.get("PubSub_Words", "channel_0")
subscriber_name = "curve"
subscriber_config_section = "PubSub_Words"
sub = ZMQ_PubSub.ZMQSub(configfile, subscriber_config_section, channel, subscriber_name)
# FUNCTIONS #
publisher.info("""Suscribed to channel {0}""".format(channel))
while True:
sub.get_and_lpush(r_serv)
if r_serv.sismember("SHUTDOWN_FLAGS", "Curve_Q"):
r_serv.srem("SHUTDOWN_FLAGS", "Curve_Q")
print "Shutdown Flag Up: Terminating"
publisher.warning("Shutdown Flag Up: Terminating.")
break
if __name__ == "__main__":
main()
publisher.port = 6380
publisher.channel = "Queuing"
config_section = 'PubSub_Words'
config_channel = 'channel_0'
subscriber_name = 'curve'
h = Helper.Redis_Queues(config_section, config_channel, subscriber_name)
h.zmq_sub(config_section)
h.redis_queue_subscribe(publisher)

View File

@ -13,61 +13,46 @@ Requirements:
"""
import redis
import ConfigParser
import os
import time
from packages import Paste
from packages import ZMQ_PubSub
from pubsublogger import publisher
from pybloomfilter import BloomFilter
configfile = './packages/config.cfg'
import Helper
if __name__ == "__main__":
publisher.port = 6380
publisher.channel = "Script"
def main():
"""Main Function"""
config_section = 'PubSub_Global'
config_channel = 'channel'
subscriber_name = 'duplicate'
# CONFIG #
cfg = ConfigParser.ConfigParser()
cfg.read(configfile)
h = Helper.Redis_Queues(config_section, config_channel, subscriber_name)
# REDIS #
# DB QUEUE ( MEMORY )
r_Q_serv = redis.StrictRedis(
host=cfg.get("Redis_Queues", "host"),
port=cfg.getint("Redis_Queues", "port"),
db=cfg.getint("Redis_Queues", "db"))
r_serv_merge = redis.StrictRedis(
host=cfg.get("Redis_Data_Merging", "host"),
port=cfg.getint("Redis_Data_Merging", "port"),
db=cfg.getint("Redis_Data_Merging", "db"))
# Subscriber
h.zmq_sub(config_section)
# REDIS #
# DB OBJECT & HASHS ( DISK )
# FIXME increase flexibility
dico_redis = {}
for year in xrange(2013, 2015):
for month in xrange(0, 16):
dico_redis[str(year)+str(month).zfill(2)] = redis.StrictRedis(
host=cfg.get("Redis_Level_DB", "host"),
port=year,
host=h.config.get("Redis_Level_DB", "host"), port=year,
db=month)
# LOGGING #
publisher.channel = "Script"
# ZMQ #
channel = cfg.get("PubSub_Global", "channel")
subscriber_name = "duplicate"
subscriber_config_section = "PubSub_Global"
sub = ZMQ_PubSub.ZMQSub(configfile, subscriber_config_section, channel, subscriber_name)
# FUNCTIONS #
publisher.info("""Script duplicate subscribed to channel {0}""".format(cfg.get("PubSub_Global", "channel")))
publisher.info("""Script duplicate subscribed to channel {0}""".format(
h.config.get("PubSub_Global", "channel")))
set_limit = 100
bloompath = os.path.join(os.environ['AIL_HOME'],
h.config.get("Directories", "bloomfilters"))
bloop_path_set = set()
while True:
try:
super_dico = {}
@ -77,15 +62,14 @@ def main():
x = time.time()
message = sub.get_msg_from_queue(r_Q_serv)
message = h.redis_rpop()
if message is not None:
path = message.split(" ", -1)[-1]
PST = Paste.Paste(path)
else:
publisher.debug("Script Attribute is idling 10s")
time.sleep(10)
if r_Q_serv.sismember("SHUTDOWN_FLAGS", "Duplicate"):
r_Q_serv.srem("SHUTDOWN_FLAGS", "Duplicate")
if h.redis_queue_shutdown():
print "Shutdown Flag Up: Terminating"
publisher.warning("Shutdown Flag Up: Terminating.")
break
@ -97,19 +81,14 @@ def main():
r_serv1 = dico_redis[PST.p_date.year + PST.p_date.month]
# Creating the bloom filter name: bloomyyyymm
bloomname = 'bloom' + PST.p_date.year + PST.p_date.month
bloompath = cfg.get("Directories", "bloomfilters")
filebloompath = bloompath + bloomname
# datetime.date(int(PST.p_date.year),int(PST.p_date.month),int(PST.p_date.day)).timetuple().tm_yday % 7
filebloompath = os.path.join(bloompath, 'bloom' + PST.p_date.year +
PST.p_date.month)
if os.path.exists(filebloompath):
bloom = BloomFilter.open(filebloompath)
else:
bloom = BloomFilter(100000000, 0.01, filebloompath)
r_Q_serv.sadd("bloomlist", filebloompath)
bloop_path_set.add(filebloompath)
# UNIQUE INDEX HASHS TABLE
r_serv0 = dico_redis["201300"]
@ -121,45 +100,43 @@ def main():
# For each bloom filter
opened_bloom = []
for bloo in r_Q_serv.smembers("bloomlist"):
for bloo in bloop_path_set:
# Opening blooms
opened_bloom.append(BloomFilter.open(bloo))
# For each hash of the paste
for hash in PST._get_hash_lines(min=5, start=1, jump=0):
for line_hash in PST._get_hash_lines(min=5, start=1, jump=0):
nb_hash_current += 1
# Adding the hash in Redis & limiting the set
if r_serv1.scard(hash) <= set_limit:
r_serv1.sadd(hash, index)
r_serv1.sadd("HASHS", hash)
if r_serv1.scard(line_hash) <= set_limit:
r_serv1.sadd(line_hash, index)
r_serv1.sadd("HASHS", line_hash)
# Adding the hash in the bloom of the month
bloom.add(hash)
bloom.add(line_hash)
# Go throught the Database of the bloom filter (of the month)
for bloo in opened_bloom:
if hash in bloo:
if line_hash in bloo:
db = bloo.name[-6:]
# Go throught the Database of the bloom filter (of the month)
# Go throught the Database of the bloom filter (month)
r_serv_bloom = dico_redis[db]
# set of index paste: set([1,2,4,65])
hash_current = r_serv_bloom.smembers(hash)
hash_current = r_serv_bloom.smembers(line_hash)
# removing itself from the list
hash_current = hash_current - set([index])
# if the hash is present at least in 1 files (already processed)
# if the hash is present at least in 1 files
# (already processed)
if len(hash_current) != 0:
hash_dico[hash] = hash_current
hash_dico[line_hash] = hash_current
# if there is data in this dictionnary
if len(hash_dico) != 0:
super_dico[index] = hash_dico
else:
# The hash is not in this bloom
pass
###########################################################################################
###########################################################################
# if there is data in this dictionnary
if len(super_dico) != 0:
@ -171,7 +148,8 @@ def main():
for p_fname in pset:
occur_dico.setdefault(p_fname, 0)
# Count how much hash is similar per file occuring in the dictionnary
# Count how much hash is similar per file occuring
# in the dictionnary
if occur_dico[p_fname] >= 0:
occur_dico[p_fname] = occur_dico[p_fname] + 1
@ -181,10 +159,11 @@ def main():
dupl.append((paste, percentage))
# Creating the object attribute and save it.
to_print = 'Duplicate;{};{};{};'.format(PST.p_source, PST.p_date, PST.p_name)
to_print = 'Duplicate;{};{};{};'.format(
PST.p_source, PST.p_date, PST.p_name)
if dupl != []:
PST.__setattr__("p_duplicate", dupl)
PST.save_attribute_redis(r_serv_merge, "p_duplicate", dupl)
PST.save_attribute_redis("p_duplicate", dupl)
publisher.info('{}Detected {}'.format(to_print, len(dupl)))
y = time.time()
@ -193,7 +172,3 @@ def main():
except IOError:
print "CRC Checksum Failed on :", PST.p_path
publisher.error('{}CRC Checksum Failed'.format(to_print))
pass
if __name__ == "__main__":
main()

View File

@ -1,45 +1,18 @@
#!/usr/bin/env python2
# -*-coding:UTF-8 -*
import redis
import ConfigParser
from packages import ZMQ_PubSub
from pubsublogger import publisher
configfile = './packages/config.cfg'
def main():
"""Main Function"""
# CONFIG #
cfg = ConfigParser.ConfigParser()
cfg.read(configfile)
# REDIS #
r_serv = redis.StrictRedis(
host=cfg.get("Redis_Queues", "host"),
port=cfg.getint("Redis_Queues", "port"),
db=cfg.getint("Redis_Queues", "db"))
# LOGGING #
publisher.channel = "Queuing"
# ZMQ #
channel = cfg.get("PubSub_Global", "channel")
sub = ZMQ_PubSub.ZMQSub(configfile, "PubSub_Global", channel, "duplicate")
# FUNCTIONS #
publisher.info("""Suscribed to channel {0}""".format(channel))
while True:
sub.get_and_lpush(r_serv)
if r_serv.sismember("SHUTDOWN_FLAGS", "Duplicate_Q"):
r_serv.srem("SHUTDOWN_FLAGS", "Duplicate_Q")
print "Shutdown Flag Up: Terminating"
publisher.warning("Shutdown Flag Up: Terminating.")
break
import Helper
if __name__ == "__main__":
main()
publisher.port = 6380
publisher.channel = 'Queuing'
config_section = 'PubSub_Global'
config_channel = 'channel'
subscriber_name = 'duplicate'
h = Helper.Redis_Queues(config_section, config_channel, subscriber_name)
h.zmq_sub(config_section)
h.redis_queue_subscribe(publisher)

View File

@ -9,38 +9,38 @@ The ZMQ_Sub_Indexer modules is fetching the list of files to be processed
and index each file with a full-text indexer (Whoosh until now).
"""
import redis
import ConfigParser
import time
from packages import Paste
from packages import ZMQ_PubSub
from pubsublogger import publisher
from whoosh.index import create_in, exists_in, open_dir
from whoosh.fields import Schema, TEXT, ID
import os
configfile = './packages/config.cfg'
import Helper
def main():
"""Main Function"""
if __name__ == "__main__":
publisher.port = 6380
publisher.channel = "Script"
# CONFIG #
cfg = ConfigParser.ConfigParser()
cfg.read(configfile)
# Subscriber
sub_config_section = 'PubSub_Global'
sub_name = 'indexer'
# Redis
r_serv1 = redis.StrictRedis(
host=cfg.get("Redis_Queues", "host"),
port=cfg.getint("Redis_Queues", "port"),
db=cfg.getint("Redis_Queues", "db"))
config_section = 'PubSub_Global'
config_channel = 'channel'
subscriber_name = 'indexer'
h = Helper.Redis_Queues(config_section, config_channel, subscriber_name)
# Indexer configuration - index dir and schema setup
indexpath = cfg.get("Indexer", "path")
indexertype = cfg.get("Indexer", "type")
indexpath = h.config.get("Indexer", "path")
indexertype = h.config.get("Indexer", "type")
if indexertype == "whoosh":
schema = Schema(title=TEXT(stored=True), path=ID(stored=True, unique=True), content=TEXT)
schema = Schema(title=TEXT(stored=True), path=ID(stored=True,
unique=True),
content=TEXT)
if not os.path.exists(indexpath):
os.mkdir(indexpath)
if not exists_in(indexpath):
@ -49,29 +49,16 @@ def main():
ix = open_dir(indexpath)
# LOGGING #
publisher.channel = "Script"
# ZMQ #
# Subscriber
channel = cfg.get("PubSub_Global", "channel")
subscriber_name = "indexer"
subscriber_config_section = "PubSub_Global"
sub = ZMQ_PubSub.ZMQSub(configfile, subscriber_config_section, channel, subscriber_name)
# FUNCTIONS #
publisher.info("""ZMQ Indexer is Running""")
while True:
try:
message = sub.get_msg_from_queue(r_serv1)
message = h.redis_rpop()
if message is not None:
PST = Paste.Paste(message.split(" ", -1)[-1])
else:
if r_serv1.sismember("SHUTDOWN_FLAGS", "Indexer"):
r_serv1.srem("SHUTDOWN_FLAGS", "Indexer")
publisher.warning("Shutdown Flag Up: Terminating.")
if h.redis_queue_shutdown():
break
publisher.debug("Script Indexer is idling 10s")
time.sleep(1)
@ -88,9 +75,5 @@ def main():
indexwriter.commit()
except IOError:
print "CRC Checksum Failed on :", PST.p_path
publisher.error('Duplicate;{};{};{};CRC Checksum Failed'.format(PST.p_source, PST.p_date, PST.p_name))
pass
if __name__ == "__main__":
main()
publisher.error('Duplicate;{};{};{};CRC Checksum Failed'.format(
PST.p_source, PST.p_date, PST.p_name))

View File

@ -12,49 +12,19 @@ handling the indexing process of the files seen.
"""
import redis
import ConfigParser
from pubsublogger import publisher
from packages import ZMQ_PubSub
configfile = './packages/config.cfg'
import Helper
def main():
"""Main Function"""
# CONFIG #
cfg = ConfigParser.ConfigParser()
cfg.read(configfile)
# REDIS #
r_serv = redis.StrictRedis(
host=cfg.get("Redis_Queues", "host"),
port=cfg.getint("Redis_Queues", "port"),
db=cfg.getint("Redis_Queues", "db"))
# LOGGING #
publisher.channel = "Queuing"
# ZMQ #
channel = cfg.get("PubSub_Global", "channel")
subscriber_name = "indexer"
sub = ZMQ_PubSub.ZMQSub(configfile, "PubSub_Global", channel, subscriber_name)
publisher.info("""Suscribed to channel {0}""".format(channel))
# Until the service is requested to be shutdown, the service
# will get the data from the global ZMQ queue and buffer it in Redis.
while True:
sub.get_and_lpush(r_serv)
if r_serv.sismember("SHUTDOWN_FLAGS", "Indexer_Q"):
r_serv.srem("SHUTDOWN_FLAGS", "Indexer_Q")
print "Shutdown Flag Up: Terminating"
publisher.warning("Shutdown Flag Up: Terminating.")
break
if __name__ == "__main__":
main()
publisher.port = 6380
publisher.channel = "Queuing"
config_section = 'PubSub_Global'
config_channel = 'channel'
subscriber_name = 'indexer'
h = Helper.Redis_Queues(config_section, config_channel, subscriber_name)
h.zmq_sub(config_section)
h.redis_queue_subscribe(publisher)

View File

@ -2,53 +2,43 @@
# -*-coding:UTF-8 -*
import redis
import ConfigParser
import pprint
import time
import dns.exception
from packages import Paste as P
from packages import Paste
from packages import lib_refine
from packages import ZMQ_PubSub
from pubsublogger import publisher
configfile = './packages/config.cfg'
import Helper
def main():
"""Main Function"""
# CONFIG #
cfg = ConfigParser.ConfigParser()
cfg.read(configfile)
# REDIS #
r_serv = redis.StrictRedis(
host=cfg.get("Redis_Queues", "host"),
port=cfg.getint("Redis_Queues", "port"),
db=cfg.getint("Redis_Queues", "db"))
r_serv1 = redis.StrictRedis(
host=cfg.get("Redis_Data_Merging", "host"),
port=cfg.getint("Redis_Data_Merging", "port"),
db=cfg.getint("Redis_Data_Merging", "db"))
r_serv2 = redis.StrictRedis(
host=cfg.get("Redis_Cache", "host"),
port=cfg.getint("Redis_Cache", "port"),
db=cfg.getint("Redis_Cache", "db"))
# LOGGING #
if __name__ == "__main__":
publisher.port = 6380
publisher.channel = "Script"
# ZMQ #
sub = ZMQ_PubSub.ZMQSub(configfile, "PubSub_Categ", "mails_categ", "emails")
config_section = 'PubSub_Categ'
config_channel = 'channel_1'
subscriber_name = 'emails'
h = Helper.Redis_Queues(config_section, config_channel, subscriber_name)
# Subscriber
h.zmq_sub(config_section)
# REDIS #
r_serv2 = redis.StrictRedis(
host=h.config.get("Redis_Cache", "host"),
port=h.config.getint("Redis_Cache", "port"),
db=h.config.getint("Redis_Cache", "db"))
# FUNCTIONS #
publisher.info("Suscribed to channel mails_categ")
message = sub.get_msg_from_queue(r_serv)
message = h.redis_rpop()
prec_filename = None
# Log as critical if there are more that that amout of valid emails
is_critical = 10
email_regex = "[a-zA-Z0-9._%+-]+@[a-zA-Z0-9.-]+\.[a-zA-Z]{2,6}"
while True:
@ -57,36 +47,35 @@ def main():
channel, filename, word, score = message.split()
if prec_filename is None or filename != prec_filename:
PST = P.Paste(filename)
MX_values = lib_refine.checking_MX_record(r_serv2, PST.get_regex(email_regex))
PST = Paste.Paste(filename)
MX_values = lib_refine.checking_MX_record(
r_serv2, PST.get_regex(email_regex))
if MX_values[0] >= 1:
PST.__setattr__(channel, MX_values)
PST.save_attribute_redis(r_serv1, channel, (MX_values[0], list(MX_values[1])))
PST.save_attribute_redis(channel, (MX_values[0],
list(MX_values[1])))
pprint.pprint(MX_values)
to_print = 'Mails;{};{};{};Checked {} e-mail(s)'.format(PST.p_source, PST.p_date, PST.p_name, MX_values[0])
if MX_values[0] > 10:
to_print = 'Mails;{};{};{};Checked {} e-mail(s)'.\
format(PST.p_source, PST.p_date, PST.p_name,
MX_values[0])
if MX_values[0] > is_critical:
publisher.warning(to_print)
else:
publisher.info(to_print)
prec_filename = filename
else:
if r_serv.sismember("SHUTDOWN_FLAGS", "Mails"):
r_serv.srem("SHUTDOWN_FLAGS", "Mails")
if h.redis_queue_shutdown():
print "Shutdown Flag Up: Terminating"
publisher.warning("Shutdown Flag Up: Terminating.")
break
publisher.debug("Script Mails is Idling 10s")
time.sleep(10)
message = sub.get_msg_from_queue(r_serv)
message = h.redis_rpop()
except dns.exception.Timeout:
# FIXME retry!
print "dns.exception.Timeout"
pass
if __name__ == "__main__":
main()

View File

@ -1,44 +1,18 @@
#!/usr/bin/env python2
# -*-coding:UTF-8 -*
import redis
import ConfigParser
from packages import ZMQ_PubSub
from pubsublogger import publisher
configfile = './packages/config.cfg'
def main():
"""Main Function"""
# CONFIG #
cfg = ConfigParser.ConfigParser()
cfg.read(configfile)
# REDIS #
r_serv = redis.StrictRedis(
host=cfg.get("Redis_Queues", "host"),
port=cfg.getint("Redis_Queues", "port"),
db=cfg.getint("Redis_Queues", "db"))
# LOGGING #
publisher.channel = "Queuing"
# ZMQ #
sub = ZMQ_PubSub.ZMQSub(configfile, "PubSub_Categ", "mails_categ", "emails")
# FUNCTIONS #
publisher.info("""Suscribed to channel {0}""".format("mails_categ"))
while True:
sub.get_and_lpush(r_serv)
if r_serv.sismember("SHUTDOWN_FLAGS", "Mails_Q"):
r_serv.srem("SHUTDOWN_FLAGS", "Mails_Q")
print "Shutdown Flag Up: Terminating"
publisher.warning("Shutdown Flag Up: Terminating.")
break
import Helper
if __name__ == "__main__":
main()
publisher.port = 6380
publisher.channel = "Queuing"
config_section = 'PubSub_Categ'
config_channel = 'channel_1'
subscriber_name = 'emails'
h = Helper.Redis_Queues(config_section, config_channel, subscriber_name)
h.zmq_sub(config_section)
h.redis_queue_subscribe(publisher)

View File

@ -6,8 +6,8 @@ The ZMQ_Sub_Onion Module
This module is consuming the Redis-list created by the ZMQ_Sub_Onion_Q Module.
It trying to extract url from paste and returning only ones which are tor related
(.onion)
It trying to extract url from paste and returning only ones which are tor
related (.onion)
..seealso:: Paste method (get_regex)
@ -21,46 +21,32 @@ Requirements
*Need the ZMQ_Sub_Onion_Q Module running to be able to work properly.
"""
import redis
import ConfigParser
import pprint
import time
from packages import Paste
from packages import ZMQ_PubSub
from pubsublogger import publisher
configfile = './packages/config.cfg'
import Helper
def main():
"""Main Function"""
# CONFIG #
cfg = ConfigParser.ConfigParser()
cfg.read(configfile)
# REDIS #
r_serv = redis.StrictRedis(
host=cfg.get("Redis_Queues", "host"),
port=cfg.getint("Redis_Queues", "port"),
db=cfg.getint("Redis_Queues", "db"))
r_serv1 = redis.StrictRedis(
host=cfg.get("Redis_Data_Merging", "host"),
port=cfg.getint("Redis_Data_Merging", "port"),
db=cfg.getint("Redis_Data_Merging", "db"))
# LOGGING #
if __name__ == "__main__":
publisher.port = 6380
publisher.channel = "Script"
# ZMQ #
Sub = ZMQ_PubSub.ZMQSub(configfile, "PubSub_Categ", "onion_categ", "tor")
config_section = 'PubSub_Categ'
config_channel = 'channel_2'
subscriber_name = 'tor'
h = Helper.Redis_Queues(config_section, config_channel, subscriber_name)
# Subscriber
h.zmq_sub(config_section)
# FUNCTIONS #
publisher.info("Script subscribed to channel onion_categ")
# Getting the first message from redis.
message = Sub.get_msg_from_queue(r_serv)
message = h.redis_rpop()
prec_filename = None
# Thanks to Faup project for this regex
@ -78,35 +64,33 @@ def main():
for x in PST.get_regex(url_regex):
# Extracting url with regex
credential, subdomain, domain, host, tld, port, resource_path, query_string, f1, f2, f3, f4 = x
credential, subdomain, domain, host, tld, port, \
resource_path, query_string, f1, f2, f3, f4 = x
if f1 == "onion":
domains_list.append(domain)
# Saving the list of extracted onion domains.
PST.__setattr__(channel, domains_list)
PST.save_attribute_redis(r_serv1, channel, domains_list)
PST.save_attribute_redis(channel, domains_list)
pprint.pprint(domains_list)
print PST.p_path
to_print = 'Onion;{};{};{};'.format(PST.p_source, PST.p_date, PST.p_name)
to_print = 'Onion;{};{};{};'.format(PST.p_source, PST.p_date,
PST.p_name)
if len(domains_list) > 0:
publisher.warning('{}Detected {} .onion(s)'.format(to_print, len(domains_list)))
publisher.warning('{}Detected {} .onion(s)'.format(
to_print, len(domains_list)))
else:
publisher.info('{}Onion related'.format(to_print))
prec_filename = filename
else:
if r_serv.sismember("SHUTDOWN_FLAGS", "Onion"):
r_serv.srem("SHUTDOWN_FLAGS", "Onion")
if h.redis_queue_shutdown():
print "Shutdown Flag Up: Terminating"
publisher.warning("Shutdown Flag Up: Terminating.")
break
publisher.debug("Script url is Idling 10s")
time.sleep(10)
message = Sub.get_msg_from_queue(r_serv)
if __name__ == "__main__":
main()
message = h.redis_rpop()

View File

@ -17,44 +17,19 @@ Requirements
*Should register to the Publisher "ZMQ_PubSub_Categ"
"""
import redis
import ConfigParser
from packages import ZMQ_PubSub
from pubsublogger import publisher
configfile = './packages/config.cfg'
import Helper
def main():
"""Main Function"""
# CONFIG #
cfg = ConfigParser.ConfigParser()
cfg.read(configfile)
# REDIS #
r_serv = redis.StrictRedis(
host=cfg.get("Redis_Queues", "host"),
port=cfg.getint("Redis_Queues", "port"),
db=cfg.getint("Redis_Queues", "db"))
# LOGGING #
publisher.channel = "Queuing"
# ZMQ #
sub = ZMQ_PubSub.ZMQSub(configfile, "PubSub_Categ", "onion_categ", "tor")
# FUNCTIONS #
publisher.info("""Suscribed to channel {0}""".format("onion_categ"))
while True:
sub.get_and_lpush(r_serv)
if r_serv.sismember("SHUTDOWN_FLAGS", "Onion_Q"):
r_serv.srem("SHUTDOWN_FLAGS", "Onion_Q")
print "Shutdown Flag Up: Terminating"
publisher.warning("Shutdown Flag Up: Terminating.")
break
if __name__ == "__main__":
main()
publisher.port = 6380
publisher.channel = "Queuing"
config_section = 'PubSub_Categ'
config_channel = 'channel_2'
subscriber_name = 'tor'
h = Helper.Redis_Queues(config_section, config_channel, subscriber_name)
h.zmq_sub(config_section)
h.redis_queue_subscribe(publisher)

View File

@ -1,13 +1,11 @@
#!/usr/bin/env python2
# -*-coding:UTF-8 -*
import redis
import ConfigParser
import pprint
import time
import dns.exception
from packages import Paste
from packages import lib_refine
from packages import ZMQ_PubSub
from pubsublogger import publisher
# Country and ASN lookup
@ -16,55 +14,39 @@ import socket
import pycountry
import ipaddress
configfile = './packages/config.cfg'
import Helper
def main():
"""Main Function"""
# CONFIG #
cfg = ConfigParser.ConfigParser()
cfg.read(configfile)
# REDIS #
r_serv = redis.StrictRedis(
host=cfg.get("Redis_Queues", "host"),
port=cfg.getint("Redis_Queues", "port"),
db=cfg.getint("Redis_Queues", "db"))
r_serv1 = redis.StrictRedis(
host=cfg.get("Redis_Data_Merging", "host"),
port=cfg.getint("Redis_Data_Merging", "port"),
db=cfg.getint("Redis_Data_Merging", "db"))
r_serv2 = redis.StrictRedis(
host=cfg.get("Redis_Cache", "host"),
port=cfg.getint("Redis_Cache", "port"),
db=cfg.getint("Redis_Cache", "db"))
# LOGGING #
if __name__ == "__main__":
publisher.port = 6380
publisher.channel = "Script"
# ZMQ #
# Subscriber
config_section = 'PubSub_Categ'
config_channel = 'channel_3'
subscriber_name = "urls"
subscriber_config_section = "PubSub_Categ"
h = Helper.Redis_Queues(config_section, config_channel, subscriber_name)
# Publisher
publisher_config_section = "PubSub_Url"
publisher_name = "adress"
pubchannel = cfg.get("PubSub_Url", "channel")
pub_config_section = "PubSub_Url"
pub_config_channel = 'channel'
h.zmq_pub(pub_config_section, pub_config_channel)
# Subscriber
h.zmq_sub(config_section)
# REDIS #
r_serv2 = redis.StrictRedis(
host=h.config.get("Redis_Cache", "host"),
port=h.config.getint("Redis_Cache", "port"),
db=h.config.getint("Redis_Cache", "db"))
# Country to log as critical
cc_critical = cfg.get("PubSub_Url", "cc_critical")
sub = ZMQ_PubSub.ZMQSub(configfile, subscriber_config_section, "web_categ", subscriber_name)
pub = ZMQ_PubSub.ZMQPub(configfile, publisher_config_section, publisher_name)
cc_critical = h.config.get("PubSub_Url", "cc_critical")
# FUNCTIONS #
publisher.info("Script URL subscribed to channel web_categ")
message = sub.get_msg_from_queue(r_serv)
message = h.redis_rpop()
prec_filename = None
url_regex = "(http|https|ftp)\://([a-zA-Z0-9\.\-]+(\:[a-zA-Z0-9\.&%\$\-]+)*@)*((25[0-5]|2[0-4][0-9]|[0-1]{1}[0-9]{2}|[1-9]{1}[0-9]{1}|[1-9])\.(25[0-5]|2[0-4][0-9]|[0-1]{1}[0-9]{2}|[1-9]{1}[0-9]{1}|[1-9]|0)\.(25[0-5]|2[0-4][0-9]|[0-1]{1}[0-9]{2}|[1-9]{1}[0-9]{1}|[1-9]|0)\.(25[0-5]|2[0-4][0-9]|[0-1]{1}[0-9]{2}|[1-9]{1}[0-9]{1}|[0-9])|localhost|([a-zA-Z0-9\-]+\.)*[a-zA-Z0-9\-]+\.(com|edu|gov|int|mil|net|org|biz|arpa|info|name|pro|aero|coop|museum|[a-zA-Z]{2}))(\:[0-9]+)*(/($|[a-zA-Z0-9\.\,\?\'\\\+&%\$#\=~_\-]+))*"
@ -79,11 +61,12 @@ def main():
PST = Paste.Paste(filename)
client = ip2asn()
for x in PST.get_regex(url_regex):
scheme, credential, subdomain, domain, host, tld, port, resource_path, query_string, f1, f2, f3, f4 = x
scheme, credential, subdomain, domain, host, tld, \
port, resource_path, query_string, f1, f2, f3, \
f4 = x
domains_list.append(domain)
msg = pubchannel + " " + str(x)
pub.send_message(msg)
publisher.debug('{0} Published'.format(x))
h.zmq_pub_send(str(x))
publisher.debug('{} Published'.format(x))
if f1 == "onion":
print domain
@ -107,34 +90,36 @@ def main():
# EU is not an official ISO 3166 code (but used by RIPE
# IP allocation)
if cc is not None and cc != "EU":
print hostl, asn, cc, pycountry.countries.get(alpha2=cc).name
print hostl, asn, cc, \
pycountry.countries.get(alpha2=cc).name
if cc == cc_critical:
publisher.warning('{0};{1};{2};{3};{4}'.format("Url", PST.p_source, PST.p_date, PST.p_name, "Detected " + hostl + " " + cc))
publisher.warning(
'Url;{};{};{};Detected {} {}'.format(
PST.p_source, PST.p_date, PST.p_name,
hostl, cc))
else:
print hostl, asn, cc
A_values = lib_refine.checking_A_record(r_serv2, domains_list)
A_values = lib_refine.checking_A_record(r_serv2,
domains_list)
if A_values[0] >= 1:
PST.__setattr__(channel, A_values)
PST.save_attribute_redis(r_serv1, channel, (A_values[0], list(A_values[1])))
PST.save_attribute_redis(channel, (A_values[0],
list(A_values[1])))
pprint.pprint(A_values)
publisher.info('{0};{1};{2};{3};{4}'.format("Url", PST.p_source, PST.p_date, PST.p_name, "Checked " + str(A_values[0]) + " URL"))
publisher.info('Url;{};{};{};Checked {} URL'.format(
PST.p_source, PST.p_date, PST.p_name, A_values[0]))
prec_filename = filename
else:
if r_serv.sismember("SHUTDOWN_FLAGS", "Urls"):
r_serv.srem("SHUTDOWN_FLAGS", "Urls")
if h.redis_queue_shutdown():
print "Shutdown Flag Up: Terminating"
publisher.warning("Shutdown Flag Up: Terminating.")
break
publisher.debug("Script url is Idling 10s")
time.sleep(10)
message = sub.get_msg_from_queue(r_serv)
message = h.redis_rpop()
except dns.exception.Timeout:
print "dns.exception.Timeout", A_values
pass
if __name__ == "__main__":
main()

View File

@ -1,44 +1,19 @@
#!/usr/bin/env python2
# -*-coding:UTF-8 -*
import redis
import ConfigParser
from packages import ZMQ_PubSub
from pubsublogger import publisher
configfile = './packages/config.cfg'
import Helper
def main():
"""Main Function"""
# CONFIG #
cfg = ConfigParser.ConfigParser()
cfg.read(configfile)
# REDIS #
r_serv = redis.StrictRedis(
host=cfg.get("Redis_Queues", "host"),
port=cfg.getint("Redis_Queues", "port"),
db=cfg.getint("Redis_Queues", "db"))
# LOGGING #
publisher.channel = "Queuing"
# ZMQ #
sub = ZMQ_PubSub.ZMQSub(configfile, "PubSub_Categ", "web_categ", "urls")
# FUNCTIONS #
publisher.info("""Suscribed to channel {0}""".format("web_categ"))
while True:
sub.get_and_lpush(r_serv)
if r_serv.sismember("SHUTDOWN_FLAGS", "Urls_Q"):
r_serv.srem("SHUTDOWN_FLAGS", "Urls_Q")
print "Shutdown Flag Up: Terminating"
publisher.warning("Shutdown Flag Up: Terminating.")
break
if __name__ == "__main__":
main()
publisher.port = 6380
publisher.channel = "Queuing"
config_section = 'PubSub_Categ'
config_channel = 'channel_3'
subscriber_name = 'urls'
h = Helper.Redis_Queues(config_section, config_channel, subscriber_name)
h.zmq_sub(config_section)
h.redis_queue_subscribe(publisher)

View File

@ -13,6 +13,7 @@
import ConfigParser
import argparse
import gzip
import os
def readdoc(path=None):
@ -21,12 +22,12 @@ def readdoc(path=None):
f = gzip.open(path, 'r')
return f.read()
configfile = './packages/config.cfg'
configfile = os.path.join(os.environ['AIL_BIN'], 'packages/config.cfg')
cfg = ConfigParser.ConfigParser()
cfg.read(configfile)
# Indexer configuration - index dir and schema setup
indexpath = cfg.get("Indexer", "path")
indexpath = os.path.join(os.environ['AIL_HOME'], cfg.get("Indexer", "path"))
indexertype = cfg.get("Indexer", "type")
argParser = argparse.ArgumentParser(description='Fulltext search for AIL')

View File

@ -19,7 +19,6 @@ Conditions to fulfill to be able to use this class correctly:
import os
import magic
import gzip
import pprint
import redis
import operator
import string
@ -44,7 +43,8 @@ class Paste(object):
This class representing a Paste as an object.
When created, the object will have by default some "main attributes"
such as the size or the date of the paste already calculated, whereas other
attributes are not set and need to be "asked to be calculated" by their methods.
attributes are not set and need to be "asked to be calculated" by their
methods.
It was design like this because some attributes take time to be calculated
such as the langage or the duplicate...
@ -56,16 +56,25 @@ class Paste(object):
def __init__(self, p_path):
configfile = './packages/config.cfg'
configfile = os.path.join(os.environ['AIL_BIN'], 'packages/config.cfg')
if not os.path.exists(configfile):
raise Exception('Unable to find the configuration file. \
Did you set environment variables? \
Or activate the virtualenv.')
cfg = ConfigParser.ConfigParser()
cfg.read(configfile)
self.cache = redis.StrictRedis(
host=cfg.get("Redis_Queues", "host"),
port=cfg.getint("Redis_Queues", "port"),
db=cfg.getint("Redis_Queues", "db"))
self.store = redis.StrictRedis(
host=cfg.get("Redis_Data_Merging", "host"),
port=cfg.getint("Redis_Data_Merging", "port"),
db=cfg.getint("Redis_Data_Merging", "db"))
self.p_path = p_path
self.p_name = self.p_path.split('/')[-1]
self.p_name = os.path.basename(self.p_path)
self.p_size = round(os.path.getsize(self.p_path)/1024.0, 2)
self.p_mime = magic.from_buffer(self.get_p_content(), mime=True)
@ -260,7 +269,7 @@ class Paste(object):
else:
return False, var
def save_all_attributes_redis(self, r_serv, key=None):
def save_all_attributes_redis(self, key=None):
"""
Saving all the attributes in a "Redis-like" Database (Redis, LevelDB)
@ -277,7 +286,7 @@ class Paste(object):
"""
# LevelDB Compatibility
p = r_serv.pipeline(False)
p = self.store.pipeline(False)
p.hset(self.p_path, "p_name", self.p_name)
p.hset(self.p_path, "p_size", self.p_size)
p.hset(self.p_path, "p_mime", self.p_mime)
@ -296,14 +305,14 @@ class Paste(object):
pass
p.execute()
def save_attribute_redis(self, r_serv, attr_name, value):
def save_attribute_redis(self, attr_name, value):
"""
Save an attribute as a field
"""
if type(value) == set:
r_serv.hset(self.p_path, attr_name, json.dumps(list(value)))
self.store.hset(self.p_path, attr_name, json.dumps(list(value)))
else:
r_serv.hset(self.p_path, attr_name, json.dumps(value))
self.store.hset(self.p_path, attr_name, json.dumps(value))
def _get_from_redis(self, r_serv):
return r_serv.hgetall(self.p_hash)
@ -366,8 +375,3 @@ class Paste(object):
if match != '' and len(match) < 100:
matchs.append(match)
return matchs
if __name__ == "__main__":
pp = pprint.PrettyPrinter(indent=4)
main()

View File

@ -1,160 +0,0 @@
#!/usr/bin/python2.7
"""
The ``ZMQ PubSub`` Modules
==========================
"""
import zmq
import ConfigParser
class PubSub(object):
"""
The PubSub class is a ``Virtual Class`` which regroup the shared attribute
of a Publisher ZeroMQ and a Subcriber ZeroMQ
:param file_conf: -- (str) The filepath of the configuration file used (.cfg)
:param log_channel: -- (str) The channel used as a log channel
:param ps_name: -- (str) The "ID" of the Publisher/Subcriber
:return: PubSub Object
..note:: The ps_name was implemented to separate subscriber queues in redis
when they are listening on a same "stream"
..seealso:: Method of the ZMQSub class
..todo:: Create Implementing a log channel as an attribute of this virtual class.
"""
def __init__(self, file_conf, log_channel, ps_name):
self._ps_name = ps_name
self._config_parser = ConfigParser.ConfigParser()
self._config_file = file_conf # "./packages/config.cfg"
self._config_parser.read(self._config_file)
self._context_zmq = zmq.Context()
# self._logging_publisher_channel = log_channel # "Default"
# publisher.channel(self._logging_publisher_channel)
class ZMQPub(PubSub):
"""
This class create a ZMQ Publisher which is able to send_message to a choosen socket.
:param pub_config_section: -- (str) The name of the section in the config file to get the settings
:return: ZMQPub Object
:Example:
Extract of the config file:
[PubSub_Categ]
adress = tcp://127.0.0.1:5003
Creating the object and sending message:
MyPublisher = ZMQPub('./packages/config.cfg', 'PubSub_Categ', 'pubcateg')
msg = "categ1"+" "+"Im the data sent on the categ1 channel"
MyPublisher.send_message(msg)
..note:: The ps_name attribute for a publisher is "optionnal" but still required to be
instantiated correctly.
"""
def __init__(self, file_conf, pub_config_section, ps_name):
super(ZMQPub, self).__init__(file_conf, "Default", ps_name)
self._pub_config_section = pub_config_section
self._pubsocket = self._context_zmq.socket(zmq.PUB)
self._pub_adress = self._config_parser.get(self._pub_config_section, "adress")
self._pubsocket.bind(self._config_parser.get(self._pub_config_section, "adress"))
def send_message(self, message):
"""Send a message throught the publisher socket"""
self._pubsocket.send(message)
class ZMQSub(PubSub):
"""
This class create a ZMQ Subcriber which is able to receive message directly or
throught redis as a buffer.
The redis buffer is usefull when the subcriber do a time consuming job which is
desynchronising it from the stream of data received.
The redis buffer ensures that no data will be loss waiting to be processed.
:param sub_config_section: -- (str) The name of the section in the config file to get the settings
:param channel: -- (str) The listening channel of the Subcriber.
:return: ZMQSub Object
:Example:
Extract of the config file:
[PubSub_Global]
adress = tcp://127.0.0.1:5000
channel = filelist
Creating the object and receiving data + pushing to redis (redis buffering):
r_serv = redis.StrictRedis(
host = 127.0.0.1,
port = 6387,
db = 0)
channel = cfg.get("PubSub_Global", "channel")
MySubscriber = ZMQSub('./packages/config.cfg',"PubSub_Global", channel, "duplicate")
while True:
MySubscriber.get_and_lpush(r_serv)
Inside another script use this line to retrive the data from redis.
...
while True:
MySubscriber.get_msg_from_queue(r_serv)
...
..note:: If you don't want any redis buffering simply use the "get_message" method
"""
def __init__(self, file_conf, sub_config_section, channel, ps_name):
super(ZMQSub, self).__init__(file_conf, "Default", ps_name)
self._sub_config_section = sub_config_section
self._subsocket = self._context_zmq.socket(zmq.SUB)
self._sub_adress = self._config_parser.get(self._sub_config_section, "adress")
self._subsocket.connect(self._config_parser.get(self._sub_config_section, "adress"))
self._channel = channel
self._subsocket.setsockopt(zmq.SUBSCRIBE, self._channel)
def get_message(self):
"""
Get the first sent message from a Publisher.
:return: (str) Message from Publisher
"""
return self._subsocket.recv()
def get_and_lpush(self, r_serv):
"""
Get the first sent message from a Publisher and storing it in redis
..note:: This function also create a set named "queue" for monitoring needs
"""
r_serv.sadd("queues", self._channel+self._ps_name)
r_serv.lpush(self._channel+self._ps_name, self._subsocket.recv())
def get_msg_from_queue(self, r_serv):
"""
Get the first sent message from a Redis List
:return: (str) Message from Publisher
"""
return r_serv.rpop(self._channel+self._ps_name)

View File

@ -1,61 +0,0 @@
[Directories]
bloomfilters = /home/user/Blooms/
pastes = /home/user/PASTES/
##### Redis #####
[Redis_Cache]
host = localhost
port = 6379
db = 0
[Redis_Log]
host = localhost
port = 6380
db = 0
[Redis_Queues]
host = localhost
port = 6381
db = 0
[Redis_Data_Merging]
host = localhost
port = 6379
db = 1
##### LevelDB #####
[Redis_Level_DB]
host = localhost
port = 2013
db = 0
[Redis_Level_DB_Hashs]
host = localhost
port = 2013
db = 1
# PUB / SUB : ZMQ
[Feed]
adress = tcp://crf.circl.lu:5556
topicfilter = 102
[PubSub_Global]
adress = tcp://127.0.0.1:5000
channel = filelist
[PubSub_Longlines]
adress = tcp://127.0.0.1:5001
channel_0 = Longlines
channel_1 = Shortlines
[PubSub_Words]
adress = tcp://127.0.0.1:5002
channel_0 = words
[PubSub_Categ]
adress = tcp://127.0.0.1:5003
#Channels are dynamic (1 channel per categ)
[PubSub_Url]
adress = tcp://127.0.0.1:5004
channel = urls

View File

@ -1,8 +1,8 @@
[Directories]
bloomfilters = /home/user/Blooms/
pastes = /home/user/PASTES/
wordtrending_csv = /home/user/AIL/var/www/static/csv/wordstrendingdata
wordsfile = /home/user/AIL/files/wordfile
bloomfilters = Blooms
pastes = PASTES
wordtrending_csv = var/www/static/csv/wordstrendingdata
wordsfile = files/wordfile
##### Redis #####
[Redis_Cache]
@ -56,7 +56,10 @@ channel_0 = words
[PubSub_Categ]
adress = tcp://127.0.0.1:5003
#Channels are dynamic (1 channel per categ)
channel_0 = creditcard_categ
channel_1 = mails_categ
channel_2 = onion_categ
channel_3 = web_categ
[PubSub_Url]
adress = tcp://127.0.0.1:5004
@ -67,4 +70,4 @@ cc_critical = DE
# Indexer configuration
[Indexer]
type = whoosh
path = /home/user/indexdir
path = indexdir

View File

@ -45,7 +45,7 @@ def create_dirfile(r_serv, directory, overwrite):
r_serv.delete("filelist")
for x in listdirectory(directory):
r_serv.rpush("filelist", x)
r_serv.lpush("filelist", x)
publisher.info("The list was overwritten")
@ -53,13 +53,13 @@ def create_dirfile(r_serv, directory, overwrite):
if r_serv.llen("filelist") == 0:
for x in listdirectory(directory):
r_serv.rpush("filelist", x)
r_serv.lpush("filelist", x)
publisher.info("New list created")
else:
for x in listdirectory(directory):
r_serv.rpush("filelist", x)
r_serv.lpush("filelist", x)
publisher.info("The list was updated with new elements")

View File

@ -38,7 +38,7 @@ daemonize no
# When running daemonized, Redis writes a pid file in /var/run/redis.pid by
# default. You can specify a custom pid file location here.
pidfile /var/run/redis.pid
#pidfile /var/run/redis.pid
# Accept connections on the specified port, default is 6379.
# If port 0 is specified Redis will not listen on a TCP socket.
@ -67,8 +67,8 @@ tcp-backlog 511
# incoming connections. There is no default, so Redis will not listen
# on a unix socket when not specified.
#
unixsocket /tmp/redis.sock
unixsocketperm 755
#unixsocket /tmp/redis.sock
#unixsocketperm 755
# Close the connection after a client is idle for N seconds (0 to disable)
timeout 0
@ -180,9 +180,9 @@ dbfilename dump.rdb
#
# The DB will be written inside this directory, with the filename specified
# above using the 'dbfilename' configuration directive.
#
#
# The Append Only File will also be created inside this directory.
#
#
# Note that you must specify a directory here, not a file name.
dir ./
@ -331,7 +331,7 @@ slave-priority 100
#
# This should stay commented out for backward compatibility and because most
# people do not need auth (e.g. they run their own servers).
#
#
# Warning: since Redis is pretty fast an outside user can try up to
# 150k passwords per second against a good box. This means that you should
# use a very strong password otherwise it will be very easy to break.
@ -397,14 +397,14 @@ slave-priority 100
# MAXMEMORY POLICY: how Redis will select what to remove when maxmemory
# is reached. You can select among five behaviors:
#
#
# volatile-lru -> remove the key with an expire set using an LRU algorithm
# allkeys-lru -> remove any key accordingly to the LRU algorithm
# volatile-random -> remove a random key with an expire set
# allkeys-random -> remove a random key, any key
# volatile-ttl -> remove the key with the nearest expire time (minor TTL)
# noeviction -> don't expire at all, just return an error on write operations
#
#
# Note: with any of the above policies, Redis will return an error on write
# operations, when there are not suitable keys for eviction.
#
@ -453,7 +453,7 @@ appendonly no
appendfilename "appendonly.aof"
# The fsync() call tells the Operating System to actually write data on disk
# instead to wait for more data in the output buffer. Some OS will really flush
# instead to wait for more data in the output buffer. Some OS will really flush
# data on disk, some other OS will just try to do it ASAP.
#
# Redis supports three different modes:
@ -494,7 +494,7 @@ appendfsync everysec
# the same as "appendfsync none". In practical terms, this means that it is
# possible to lose up to 30 seconds of log in the worst scenario (with the
# default Linux settings).
#
#
# If you have latency problems turn this to "yes". Otherwise leave it as
# "no" that is the safest pick from the point of view of durability.
@ -503,7 +503,7 @@ no-appendfsync-on-rewrite no
# Automatic rewrite of the append only file.
# Redis is able to automatically rewrite the log file implicitly calling
# BGREWRITEAOF when the AOF log size grows by the specified percentage.
#
#
# This is how it works: Redis remembers the size of the AOF file after the
# latest rewrite (if no rewrite has happened since the restart, the size of
# the AOF at startup is used).
@ -546,7 +546,7 @@ lua-time-limit 5000
# but just the time needed to actually execute the command (this is the only
# stage of command execution where the thread is blocked and can not serve
# other requests in the meantime).
#
#
# You can configure the slow log with two parameters: one tells Redis
# what is the execution time, in microseconds, to exceed in order for the
# command to get logged, and the other parameter is the length of the
@ -566,7 +566,7 @@ slowlog-max-len 128
# Redis can notify Pub/Sub clients about events happening in the key space.
# This feature is documented at http://redis.io/topics/keyspace-events
#
#
# For instance if keyspace events notification is enabled, and a client
# performs a DEL operation on key "foo" stored in the Database 0, two
# messages will be published via Pub/Sub:
@ -642,7 +642,7 @@ zset-max-ziplist-value 64
# that is rehashing, the more rehashing "steps" are performed, so if the
# server is idle the rehashing is never complete and some more memory is used
# by the hash table.
#
#
# The default is to use this millisecond 10 times every second in order to
# active rehashing the main dictionaries, freeing memory when possible.
#

View File

@ -38,7 +38,7 @@ daemonize no
# When running daemonized, Redis writes a pid file in /var/run/redis.pid by
# default. You can specify a custom pid file location here.
pidfile /var/run/redis.pid
#pidfile /var/run/redis.pid
# Accept connections on the specified port, default is 6379.
# If port 0 is specified Redis will not listen on a TCP socket.
@ -67,8 +67,8 @@ tcp-backlog 511
# incoming connections. There is no default, so Redis will not listen
# on a unix socket when not specified.
#
unixsocket /tmp/redis.sock
unixsocketperm 755
#unixsocket /tmp/redis.sock
#unixsocketperm 755
# Close the connection after a client is idle for N seconds (0 to disable)
timeout 0
@ -180,9 +180,9 @@ dbfilename dump2.rdb
#
# The DB will be written inside this directory, with the filename specified
# above using the 'dbfilename' configuration directive.
#
#
# The Append Only File will also be created inside this directory.
#
#
# Note that you must specify a directory here, not a file name.
dir ./
@ -331,7 +331,7 @@ slave-priority 100
#
# This should stay commented out for backward compatibility and because most
# people do not need auth (e.g. they run their own servers).
#
#
# Warning: since Redis is pretty fast an outside user can try up to
# 150k passwords per second against a good box. This means that you should
# use a very strong password otherwise it will be very easy to break.
@ -397,14 +397,14 @@ slave-priority 100
# MAXMEMORY POLICY: how Redis will select what to remove when maxmemory
# is reached. You can select among five behaviors:
#
#
# volatile-lru -> remove the key with an expire set using an LRU algorithm
# allkeys-lru -> remove any key accordingly to the LRU algorithm
# volatile-random -> remove a random key with an expire set
# allkeys-random -> remove a random key, any key
# volatile-ttl -> remove the key with the nearest expire time (minor TTL)
# noeviction -> don't expire at all, just return an error on write operations
#
#
# Note: with any of the above policies, Redis will return an error on write
# operations, when there are not suitable keys for eviction.
#
@ -453,7 +453,7 @@ appendonly no
appendfilename "appendonly.aof"
# The fsync() call tells the Operating System to actually write data on disk
# instead to wait for more data in the output buffer. Some OS will really flush
# instead to wait for more data in the output buffer. Some OS will really flush
# data on disk, some other OS will just try to do it ASAP.
#
# Redis supports three different modes:
@ -494,7 +494,7 @@ appendfsync everysec
# the same as "appendfsync none". In practical terms, this means that it is
# possible to lose up to 30 seconds of log in the worst scenario (with the
# default Linux settings).
#
#
# If you have latency problems turn this to "yes". Otherwise leave it as
# "no" that is the safest pick from the point of view of durability.
@ -503,7 +503,7 @@ no-appendfsync-on-rewrite no
# Automatic rewrite of the append only file.
# Redis is able to automatically rewrite the log file implicitly calling
# BGREWRITEAOF when the AOF log size grows by the specified percentage.
#
#
# This is how it works: Redis remembers the size of the AOF file after the
# latest rewrite (if no rewrite has happened since the restart, the size of
# the AOF at startup is used).
@ -546,7 +546,7 @@ lua-time-limit 5000
# but just the time needed to actually execute the command (this is the only
# stage of command execution where the thread is blocked and can not serve
# other requests in the meantime).
#
#
# You can configure the slow log with two parameters: one tells Redis
# what is the execution time, in microseconds, to exceed in order for the
# command to get logged, and the other parameter is the length of the
@ -566,7 +566,7 @@ slowlog-max-len 128
# Redis can notify Pub/Sub clients about events happening in the key space.
# This feature is documented at http://redis.io/topics/keyspace-events
#
#
# For instance if keyspace events notification is enabled, and a client
# performs a DEL operation on key "foo" stored in the Database 0, two
# messages will be published via Pub/Sub:
@ -642,7 +642,7 @@ zset-max-ziplist-value 64
# that is rehashing, the more rehashing "steps" are performed, so if the
# server is idle the rehashing is never complete and some more memory is used
# by the hash table.
#
#
# The default is to use this millisecond 10 times every second in order to
# active rehashing the main dictionaries, freeing memory when possible.
#

View File

@ -38,7 +38,7 @@ daemonize no
# When running daemonized, Redis writes a pid file in /var/run/redis.pid by
# default. You can specify a custom pid file location here.
pidfile /var/run/redis.pid
#pidfile /var/run/redis.pid
# Accept connections on the specified port, default is 6379.
# If port 0 is specified Redis will not listen on a TCP socket.
@ -67,8 +67,8 @@ tcp-backlog 511
# incoming connections. There is no default, so Redis will not listen
# on a unix socket when not specified.
#
unixsocket /tmp/redis.sock
unixsocketperm 755
#unixsocket /tmp/redis.sock
#unixsocketperm 755
# Close the connection after a client is idle for N seconds (0 to disable)
timeout 0
@ -180,9 +180,9 @@ dbfilename dump3.rdb
#
# The DB will be written inside this directory, with the filename specified
# above using the 'dbfilename' configuration directive.
#
#
# The Append Only File will also be created inside this directory.
#
#
# Note that you must specify a directory here, not a file name.
dir ./
################################# REPLICATION #################################
@ -330,7 +330,7 @@ slave-priority 100
#
# This should stay commented out for backward compatibility and because most
# people do not need auth (e.g. they run their own servers).
#
#
# Warning: since Redis is pretty fast an outside user can try up to
# 150k passwords per second against a good box. This means that you should
# use a very strong password otherwise it will be very easy to break.
@ -396,14 +396,14 @@ slave-priority 100
# MAXMEMORY POLICY: how Redis will select what to remove when maxmemory
# is reached. You can select among five behaviors:
#
#
# volatile-lru -> remove the key with an expire set using an LRU algorithm
# allkeys-lru -> remove any key accordingly to the LRU algorithm
# volatile-random -> remove a random key with an expire set
# allkeys-random -> remove a random key, any key
# volatile-ttl -> remove the key with the nearest expire time (minor TTL)
# noeviction -> don't expire at all, just return an error on write operations
#
#
# Note: with any of the above policies, Redis will return an error on write
# operations, when there are not suitable keys for eviction.
#
@ -452,7 +452,7 @@ appendonly no
appendfilename "appendonly.aof"
# The fsync() call tells the Operating System to actually write data on disk
# instead to wait for more data in the output buffer. Some OS will really flush
# instead to wait for more data in the output buffer. Some OS will really flush
# data on disk, some other OS will just try to do it ASAP.
#
# Redis supports three different modes:
@ -493,7 +493,7 @@ appendfsync everysec
# the same as "appendfsync none". In practical terms, this means that it is
# possible to lose up to 30 seconds of log in the worst scenario (with the
# default Linux settings).
#
#
# If you have latency problems turn this to "yes". Otherwise leave it as
# "no" that is the safest pick from the point of view of durability.
@ -502,7 +502,7 @@ no-appendfsync-on-rewrite no
# Automatic rewrite of the append only file.
# Redis is able to automatically rewrite the log file implicitly calling
# BGREWRITEAOF when the AOF log size grows by the specified percentage.
#
#
# This is how it works: Redis remembers the size of the AOF file after the
# latest rewrite (if no rewrite has happened since the restart, the size of
# the AOF at startup is used).
@ -545,7 +545,7 @@ lua-time-limit 5000
# but just the time needed to actually execute the command (this is the only
# stage of command execution where the thread is blocked and can not serve
# other requests in the meantime).
#
#
# You can configure the slow log with two parameters: one tells Redis
# what is the execution time, in microseconds, to exceed in order for the
# command to get logged, and the other parameter is the length of the
@ -565,7 +565,7 @@ slowlog-max-len 128
# Redis can notify Pub/Sub clients about events happening in the key space.
# This feature is documented at http://redis.io/topics/keyspace-events
#
#
# For instance if keyspace events notification is enabled, and a client
# performs a DEL operation on key "foo" stored in the Database 0, two
# messages will be published via Pub/Sub:
@ -641,7 +641,7 @@ zset-max-ziplist-value 64
# that is rehashing, the more rehashing "steps" are performed, so if the
# server is idle the rehashing is never complete and some more memory is used
# by the hash table.
#
#
# The default is to use this millisecond 10 times every second in order to
# active rehashing the main dictionaries, freeing memory when possible.
#

View File

@ -5,15 +5,8 @@ set -x
sudo apt-get update
sudo apt-get install python-pip python-virtualenv python-dev libfreetype6-dev screen
virtualenv AILENV
echo export AIL_HOME=$(pwd) >> ./AILENV/bin/activate
echo export AIL_BIN=$(pwd)/bin/ >> ./AILENV/bin/activate
echo export AIL_FLASK=$(pwd)/var/www/ >> ./AILENV/bin/activate
sudo apt-get install g++ python-tk
sudo apt-get install python-pip python-virtualenv python-dev libfreetype6-dev \
screen g++ python-tk
#Needed for bloom filters
sudo apt-get install libssl-dev libfreetype6-dev python-numpy
@ -24,31 +17,35 @@ sudo apt-get install libadns1 libadns1-dev
#needed for mathplotlib
test ! -L /usr/include/ft2build.h && sudo ln -s freetype2/ft2build.h /usr/include/
. ./AILENV/bin/activate
pip install -r pip_packages_requirement.txt --upgrade
pip install -U textblob
python -m textblob.download_corpora
# REDIS #
test ! -d redis/ && git clone https://github.com/antirez/redis.git
pushd redis/
git checkout 2.8
git pull
make
popd
echo export AIL_REDIS=$(pwd)/src/ >> ./AILENV/bin/activate
# REDIS LEVEL DB #
test ! -d redis-leveldb/ && git clone https://github.com/KDr2/redis-leveldb.git
pushd redis-leveldb/
git submodule init
git submodule update
make
popd
mkdir -p $AIL_HOME/{PASTES,Blooms,dumps}
virtualenv AILENV
echo export AIL_HOME=$(pwd) >> ./AILENV/bin/activate
echo export AIL_BIN=$(pwd)/bin/ >> ./AILENV/bin/activate
echo export AIL_FLASK=$(pwd)/var/www/ >> ./AILENV/bin/activate
echo export AIL_REDIS=$(pwd)/redis/src/ >> ./AILENV/bin/activate
echo export AIL_LEVELDB=$(pwd)/redis-leveldb/ >> ./AILENV/bin/activate
. ./AILENV/bin/activate
mkdir -p $AIL_HOME/{PASTES,Blooms,dumps}
mkdir -p $AIL_HOME/LEVEL_DB_DATA/{2014,2013}
pip install -r pip_packages_requirement.txt
# Download the necessary NLTK corpora
HOME=$(pwd) python -m textblob.download_corpora

View File

@ -4,7 +4,7 @@ pyzmq
dnspython
logbook
pubsublogger
textblob
#Graph
numpy
@ -29,6 +29,9 @@ texttable
#Indexer requirements
whoosh
ipaddress
pycountry
#ASN lookup requirements
http://adns-python.googlecode.com/files/adns-python-1.2.1.tar.gz
https://github.com/trolldbois/python-cymru-services/archive/master.zip