Thanks a lot for this thread.

I was starting to plan a system where multiples processes can write to a file and completely forgot that syslog is designed to do that.

It is a perfect solution, I only had to configure a local facility to receive the data and add 3 lines of code to the program (including the import).

On 28/07/2022 12:01, Paul Kudla (SCOM.CA Internet Services Inc.) wrote:

Hi - I use this python script to capture a socket (ie the log file) and then send it to syslog, i use this for all the systems that do not really support syslogging (apache etc)

basic useage

/usr/bin/nohup /programs/common/capture -s /usr/local/apache2/logs/httpd-access.log -l httpd -d 10.228.0.6:514 -p httpd & > /dev/null

i typically run this at startup in rc.local

hope this helps :

---

## cat capture #!/usr/local/bin/python3 # -*- coding: UTF-8 -*-

import os,sys,socket import datetime,time from optparse import OptionParser

from lib import *

USAGE_TEXT = '''
usage: %%prog %s[options] '''

parser = OptionParser(usage=USAGE_TEXT % '', version='0.4')

parser.add_option("-s", "--socket", dest="socket_file", help="Socket File to Capture") parser.add_option("-l", "--label", dest="label", help="Syslog Label to Insert") parser.add_option("-d", "--destination", dest="destination", help="Syslog Destibnation Server:Port") parser.add_option("-p", "--pid", dest="pid", help="PID Process Name") #parser.add_option("-e", "--email", dest="email", help="Additional Email To") #parser.add_option("-t", "--temp", dest="tempdir", help="Local Temp Directory")

options, args = parser.parse_args()

print (options.socket_file) print (options.label) print (options.destination) print (options.pid)

if options.socket_file == None : print ('Missing Socket File Information') sys.exit()

if options.label == None : print ('Missing Syslog Label Information') sys.exit()

if options.destination == None : print ('Missing Syslog Destination host:[port]') sys.exit()

if options.pid == None : print ('Missing Syslog Pid Process Name') sys.exit()

#try local syslog (/var/run/log)

UDP_IP = options.destination.split(':')

if len(UDP_IP) == 2 : #Set Port UDP_PORT = int(UDP_IP[1]) else : UDP_PORT = 514 #Default

UDP_IP = UDP_IP[0]                      #Server

#MESSAGE = str("<22>Mar 27 04:16:16 es-scom[12345] offsite.scom.ca su: Hello, World!") #MESSAGE = str("<183>Mar 27 16:17:41 scom-live[72178]: Hello World")

print("UDP target IP: %s" % UDP_IP) print("UDP target port: %s" % UDP_PORT) #print("message: %s" % MESSAGE)

count = 10

#sock = socket.socket(socket.AF_INET, socket.SOCK_DGRAM) #sock.sendto(MESSAGE, (UDP_IP, UDP_PORT)) #sock.sendto(bytes(MESSAGE, "utf-8"), (UDP_IP, UDP_PORT)) #sock.close() #sys.exit()

#def read_commands(): try: print ("Creating read pipe... %s"  %options.socket_file ) os.mkfifo(options.socket_file)    # Create pipe print ("Pipe %s created!" %options.socket_file ) except: print ("Pipe %s already exists" %options.socket_file )

#chmod 777 the file so everyone can talk to it os.system('/bin/chmod 777 %s' %options.socket_file)

with open(options.socket_file, "r") as pipecmd: while True: time.sleep(.001) try: line = pipecmd.readline() if line != '' : #New Data if line == '\n' : continue print ('Raw Text : %s' %line) encoded_string = line.encode("ascii", "ignore") line = encoded_string.decode() line = create_ascii(line) line = line.ascii print ('Line after ASCII : %s' %line) print ( 'Line Count : %s' %len(line) ) #line = data #go get my pid pid_process = '0' if options.pid == 'postfix' : #its a diverted postfix process get the actual pid from raw text pid_process = line.split('[',1)[1].split(']',1)[0]

                                else : command = commands('/bin/ps -axww | /usr/bin/grep %s' %options.pid) print () #print (command.output)

                                        for n in range (0,len(command.output)) : if '/bin/ps -axww | /usr/bin/grep' not in command.output[n] and '/usr/bin/grep' not in command.output[n] and '/usr/local/bin/python3' not in command.output[n]  : pid_process = ( command.output.split(' ')[0] ) #whats left should be my process ? break

                                print ('PID Process : %s ' %pid_process )

                                if options.destination == 'local' : #Send to log here print ('Sending to Local Syslog') log = open ('/var/run/log','w') log.write ('hello') log.close() sys.exit()

                                else : #Send via socket #Make the line in freebsd syslog format MESSAGE = '<' + str(count) + '>' + str( time.strftime("%b %d %H:%M:%S ") ) + str(options.label) + '[' + str(pid_process) + ']: ' + str(line) print ('Sent : %s' %MESSAGE ) count = count + 1 if count > 255 : count = 10

                                        # send to udp logger port specified sock = socket.socket(socket.AF_INET, socket.SOCK_DGRAM) sock.sendto(bytes(MESSAGE, "utf-8"), (UDP_IP, UDP_PORT)) sock.close()

                        else : #No data pass

                except Exception as e: exc_type, exc_obj, exc_tb = sys.exc_info() fname = os.path.split(exc_tb.tb_frame.f_code.co_filename)[1] e = str(e) + '\n\n' + str(exc_type) + '\n' + str(fname) + '\n' + str(exc_tb.tb_lineno)

                        print ('\n\nCaught Exception : %s' %e )

                        print ("Could not read cmd pipe, skipping ...")

sys.exit()

---

and lib.py

---

## cat lib3.py #Python Library written by paul kudla (c) 2011

#Load the librarys for the system

import os,sys,time,socket import string from ftplib import FTP from decimal import * from datetime import date import datetime import smtplib from email.mime.multipart import MIMEMultipart from email.mime.base import MIMEBase from email.mime.text import MIMEText from email.utils import COMMASPACE, formatdate from email import encoders import subprocess

getcontext().prec = 20

class commands: def __init__(self,command) : self.command = command #print (self.command) self.output = 'Error' self.status = '255'

                #sample #rc, gopath = subprocess.getstatusoutput('ls -a')

                self.status, self.output = subprocess.getstatusoutput(self.command)

                try: self.cr = self.output.split('\n') except : self.cr = [] try: self.count = len(self.cr) except : self.count = 0

                self.status = int(self.status)

                #return count=number of lines, cr = lines split, getoutput = actual output returned, status = return code

                return

#Email with attachment class sendmail: def __init__(self, send_from, send_to, send_subject, send_text, send_files): #send_from, send_to, send_subject, send_text, send_files): #print ('lib.py sending email') assert type(send_to)==list assert type(send_files)==list

                msg = MIMEMultipart() msg['From'] = send_from msg['To'] = COMMASPACE.join(send_to) msg['Date'] = formatdate(localtime=True) msg['Subject'] = send_subject

                msg.attach( MIMEText(send_text) )

                for f in send_files: part = MIMEBase('application', "octet-stream") part.set_payload( open(f,"rb").read() ) Encoders.encode_base64(part) part.add_header('Content-Disposition', 'attachment; filename="%s"' % os.path.basename(f)) msg.attach(part)

                try : #Send Local? smtp = smtplib.SMTP('mail.local.scom.ca') #smtp.login('backup@scom.ca','522577') #print ('Sending Email to : %s' %send_to) smtp.sendmail(send_from, send_to, msg.as_string()) smtp.close()

                except : smtp = smtplib.SMTP('mail.scom.ca') smtp.login('backup@scom.ca','522577') #print ('Sending Email to : %s' %send_to) smtp.sendmail(send_from, send_to, msg.as_string()) smtp.close()

class getdatetime: def __init__(self): self.datetime = datetime.date.today() self.datetime_now = datetime.datetime.now() self.date = str( time.strftime("%Y-%m-%d %H:%M:%S") ) self.date_long = str( time.strftime("%Y-%m-%d %H:%M:%S") ) self.date_short = str( time.strftime("%Y-%m-%d") ) self.time =  str( time.strftime("%H:%M:%S") ) self.date_time_sec = self.datetime_now.strftime ("%Y-%m-%d %H:%M:%S.%f")

#Return edi senddate string (short) 2011-10-31 into 111031

class create_ascii : def __init__(self,string_data) : self.string_data = str(string_data) import string self.printable = set(string.printable) self.list = list(filter(lambda x: x in self.printable, self.string_data)) #print (self.list)

                self.ascii = '' for n in range (0,len(self.list)) : self.ascii = self.ascii + self.list[n] self.ascii = str(self.ascii)

                return

class edi_send_date_short: def __init__(self, senddate): self.date = senddate self.result = self.date[2] + self.date[3] + self.date[5] + self.date[6] + self.date[8] + self.date[9]

        def __str__(self): return '%s' % self.result

##Return edi senddate string (long) 2011-10-31 into 20111031 class edi_send_date_long: def __init__(self, senddate): self.date = senddate self.result1 = self.date[0] + self.date[1] + self.date[2] + self.date[3] + self.date[5] + self.date[6] + self.date[8] + self.date[9] self.result2 = self.date[2] + self.date[3] + self.date[5] + self.date[6] + self.date[8] + self.date[9]

        def __str__(self): return '%s' % (self.result1,self.result2)

class gpsdeg: def __init__(self, dms): self.dms = dms self.is_positive = self.dms >= 0 self.dms = abs(self.dms) self.minutes,self.seconds = divmod(self.dms*3600,60) self.degrees,self.minutes = divmod(self.minutes,60) self.degrees = self.degrees if self.is_positive else -self.degrees

        def __str__(self): return '%s' % (self.degrees,self.minutes,self.seconds)

class degdir: def __init__(self, degrees): self.direction_data = ['N','348.75','11.25','NNE', '11.25','33.75','NE','33.75','56.25','ENE', '56.25','78.75','E','78.75','101.25','ESE','101.25','123.75','SE','123.75','146.25','SSE','146.25','168.75','S','168.75','191.25','SSW','191.25','213.75','SW','213.75','236.25','WSW','236.25','258.75','W','258.75','281.25','WNW','281.25','303.75','NW','303.75','326.25','NNW','326.25','348.75']

        def __str__(self): return '%s' % (self.direction)

class gettime: def __init__(self): self.uu = time.localtime()

                self.todaystime = str(self.uu[3]) #get the hr

                if int(self.uu[3]) < 10: #add a zero self.todaystime = '0' + self.todaystime if int(self.uu[4]) < 10: #add a zero in front self.todaystime = self.todaystime +":0"+str(self.uu[4]) else: self.todaystime = self.todaystime +":"+str(self.uu[4])

        def __str__(self): return self.todaystime

class array2dbstring: def __init__(self,array): self.data = array for self.nn in range(0,len(self.data)): print ('Data %s \t\t %s' % (str(self.data[self.nn]),str( type(self.data[self.nn])) ) ) #change all data into strings self.a = type(self.data[self.nn]) self.a = str(self.a) if 'Decimal' in self.a : self.n = str(self.data[self.nn]) #self.n = self.n.lstrip("'") #self.n = self.n.rstrip("'") #self.data[self.nn] = float(self.data[self.nn]) self.data[self.nn] = str('0.00') print (self.n)

                        if 'NoneType' in self.a : self.data[self.nn] = '' if 'datetime.datetime' in self.a : #self.data[self.nn] = str(self.data[self.nn]) #self.data[self.nn].replace self.data[self.nn] = '2012-01-25 00:00:00' self.data = str(self.data) self.data = self.data.lstrip('[') self.data = self.data.rstrip(']') self.data = self.data.replace("'NULL'","NULL") #self.data = self.data.replace(" '',", ",") #self.data = self.data.replace(" '0.00'","'100'")

        def __str__(self): return self.data

class get_hostname: def __init__(self): self.hostname = socket.gethostname()

---

Happy Thursday !!! Thanks - paul

Paul Kudla

Scom.ca Internet Services http://www.scom.ca 004-1009 Byron Street South Whitby, Ontario - Canada L1N 4S3

Toronto 416.642.7266 Main 1.866.411.7266 Fax 1.888.892.7266 Email paul@scom.ca

On 7/28/2022 6:17 AM, dovecot-bounces@dovecot.org wrote:

...

Hi, I'm searching for a possibility to have the rawlog feature in lua, which would be much easier for processing. Currently Dovecot, when activating rawlog for a user, writes everything to disk (which creates I/O), and I have to somehow read it from there. That's a bit complicated, because I have to get notified via inotify or similar when there are new files created, and then I have to start a "tail" or "epoll" mechanism on the files to get the contents in more or less real time (IMAP sessions can be multiple hours or days). It would be much easier to hook to the "raw request and response events" inside Dovecot and have the rawlog-data in a lua script, where I can prepare it and send it to another maschine for monitoring/collection/analysis/statistics or similar, for example via HTTP. Having the rawlog data available in lua would make things a lot easier. Is there any possibility at the moment to create a lua script and "hook" to those "request and response events"? If not, would it be possible to add that feature in the future? Kind regards Michael

Hi Paul, I don't understand how to use your idea/script together with the rawlog feature of Dovecot. The rawlog feature in Dovecot writes multiple files (two for each connection, one for raw requests and one for raw responses) into a predefined directory for the user. This generates dozens or hundreds of files per user per day, each file with a timestamp in it, so the filename is not predictable. How should I create "a socket" for that to capture the file contents if I don't know the filenames that will be used? Michael

Am 28-Jul-2022 13:02:16 +0200 schrieb paul@scom.ca:

Hi - I use this python script to capture a socket (ie the log file) and then send it to syslog, i use this for all the systems that do not really support syslogging (apache etc)

basic useage

/usr/bin/nohup /programs/common/capture -s /usr/local/apache2/logs/httpd-access.log -l httpd -d 10.228.0.6:514 -p httpd & > /dev/null

i typically run this at startup in rc.local

hope this helps

Hi, as far as I know I cannot configure Dovecot to pipe the rawlog into rsyslog. Or can I, how? The rawlog feature in Dovecot writes multiple files (two for each connection, one for raw requests and one for raw responses) into a predefined directory for the user. This generates dozens or hundreds of files per user per day, each file with a timestamp in it, so the filename is not predictable. Even if it works, I'm not sure if syslog (rsyslog or syslog-ng) should be (ab)used to collect the rawlog file contents, which might be hundreds of MB per minute if someone FETCHes all his emails while setting up a new account in Thunderbird or so. That sounds like a suboptional idea. Syslog cannot handle binary text I guess, and it might have limits like "line length limits" or similar. It sounds like the wrong tool for the job. Michael

Am 28-Jul-2022 15:28:16 +0200 schrieb dovecot@ptld.com:

I'm searching for a possibility to have the rawlog feature in lua, which would be much easier for processing.

It would be much easier to hook to the "raw request and response events" inside Dovecot and have the rawlog-data in a lua script, where I can prepare it and send it to another maschine for monitoring/collection/analysis/statistics or similar, for example via HTTP.

rsyslog has this feature (omprog) allowing you to setup any script/program for it to pipe logs to in real time.

https://www.rsyslog.com/doc/master/configuration/modules/omprog.html https://github.com/rsyslog/rsyslog/blob/master/plugins/external/INTERFACE.md

Works similar in concept to postfix policy servers if you are familiar with them.

Hi Aki,

On 08.08.22 13:54, Aki Tuomi wrote:

Hi, Michael, did you consider my suggestion to use raw events instead of rawlogs for this?

I was writing an answer to you next :-)

As far as I can see, the "Event Export" only exports events of the requests, but not the full raw responses, correct? https://doc.dovecot.org/configuration_manual/event_export/

I need the complete rawlog that currently is written to the rawlog directory, which means the raw requests (IMAP, POP3 commands), and the raw response lines (for example a FETCH response -> a 20 MB mail content). Everything that could be seen on the wire via tcpdump after authentication (which is the rawlog of Dovecot as far as I can see).

I need the rawlog feature, but not written to multiple files (which I have to collect in realtime with some black magic), but for example in a lua-script, which would make it a lot easier to analyse and/or send it to an HTTP endpoint.

Maybe there are other possibilities, for example sending the rawlog of a user to a single file (or pipe/socket), where I can easily receive the raw logs for that user and send it to an HTTP endpoint. That's a lot easier than to "watch" a directory for new files, detect changes to existing files, collect them and send them via HTTP.

I somehow need to send the raw log of specific users in realtime (maybe with a few seconds delay) to an HTTP endpoint (where each request or response is a single HTTP request, maybe we could also batch some requests and responses to reduce the HTTP requests to the endpoint).

The current implementation of the rawlog feature is nice for manually debugging a single user, but when debugging/monitoring multiple users automatically, collect the logs and send them to a central place, it's hard to use ("watching" directories for changes via inotify, and run "tail" on the files for hours and days is not fun and can easily break).

Michael

...

On 08/08/2022 14:52 EESTmichael.zork@feierfighter.de wrote:

Hi,

as far as I know I cannot configure Dovecot to pipe the rawlog into rsyslog. Or can I, how?

The rawlog feature in Dovecot writes multiple files (two for each connection, one for raw requests and one for raw responses) into a predefined directory for the user. This generates dozens or hundreds of files per user per day, each file with a timestamp in it, so the filename is not predictable.

Even if it works, I'm not sure if syslog (rsyslog or syslog-ng) should be (ab)used to collect the rawlog file contents, which might be hundreds of MB per minute if someone FETCHes all his emails while setting up a new account in Thunderbird or so. That sounds like a suboptional idea. Syslog cannot handle binary text I guess, and it might have limits like "line length limits" or similar. It sounds like the wrong tool for the job.

Michael

Am 28-Jul-2022 15:28:16 +0200 schriebdovecot@ptld.com:

...

...

I'm searching for a possibility to have the rawlog feature in lua, which would be much easier for processing.

It would be much easier to hook to the "raw request and response events" inside Dovecot and have the rawlog-data in a lua script, where I can prepare it and send it to another maschine for monitoring/collection/analysis/statistics or similar, for example via HTTP. rsyslog has this feature (omprog) allowing you to setup any script/program for it to pipe logs to in real time.

https://www.rsyslog.com/doc/master/configuration/modules/omprog.html https://github.com/rsyslog/rsyslog/blob/master/plugins/external/INTERFACE.md

Works similar in concept to postfix policy servers if you are familiar with them.