[Python-es] Lectura de archivos planos tipo logsurfer
Esteban Dauksis
esteban en dauksis.com
Lun Abr 18 17:20:21 CEST 2011
Os lo reenvío, espero que ésta vez llegue a la lista.
Saludos
Hola, hace algún tiempo me embarqué en algo parecido.... algo que no
te han comentado de momento es gestionar la rotación/truncado de
archivos de log... lo mejor que encontré y que no recuerdo de dónde lo
saqué es lo siguiente:
# -*- coding: utf-8 -*-
"""
Module to allow for reading lines from a continuously-growing file (such as
a system log). Handles log files that get rotated/trucated out from under
us. Inspired by the Perl File::Tail module.
Example:
t = filetail.Tail("log.txt")
while True:
line = t.nextline()
# do something with the line
or:
t = filetail.Tail("log.txt")
for line in t:
# do something
pass
"""
from os import stat
from os.path import abspath
from stat import ST_SIZE
from time import sleep, time
class Tail(object):
"""The Tail monitor object."""
def __init__(self, path, only_new = False,
min_sleep = 1,
sleep_interval = 1,
max_sleep = 60):
"""Initialize a tail monitor.
path: filename to open
only_new: By default, the tail monitor will start reading from
the beginning of the file when first opened. Set only_new to
True to have it skip to the end when it first opens, so that
you only get the new additions that arrive after you start
monitoring.
min_sleep: Shortest interval in seconds to sleep when waiting
for more input to arrive. Defaults to 1.0 second.
sleep_interval: The tail monitor will dynamically recompute an
appropriate sleep interval based on a sliding window of data
arrival rate. You can set sleep_interval here to seed it
initially if the default of 1.0 second doesn't work for you
and you don't want to wait for it to converge.
max_sleep: Maximum interval in seconds to sleep when waiting
for more input to arrive. Also, if this many seconds have
elapsed without getting any new data, the tail monitor will
check to see if the log got truncated (rotated) and will
quietly reopen itself if this was the case. Defaults to 60.0
seconds.
"""
# remember path to file in case I need to reopen
self.path = abspath(path)
self.f = open(self.path,"r")
self.min_sleep = min_sleep * 1.0
self.sleep_interval = sleep_interval * 1.0
self.max_sleep = max_sleep * 1.0
if only_new:
# seek to current end of file
file_len = stat(path)[ST_SIZE]
self.f.seek(file_len)
self.pos = self.f.tell() # where am I in the file?
self.last_read = time() # when did I last get some data?
self.queue = [] # queue of lines that are ready
self.window = [] # sliding window for dynamically
# adjusting the sleep_interval
def _recompute_rate(self, n, start, stop):
"""Internal function for recomputing the sleep interval. I get
called with a number of lines that appeared between the start and
stop times; this will get added to a sliding window, and I will
recompute the average interarrival rate over the last window.
"""
self.window.append((n, start, stop))
purge_idx = -1 # index of the highest old record
tot_n = 0 # total arrivals in the window
tot_start = stop # earliest time in the window
tot_stop = start # latest time in the window
for i, record in enumerate(self.window):
(i_n, i_start, i_stop) = record
if i_stop < start - self.max_sleep:
# window size is based on self.max_sleep; this record has
# fallen out of the window
purge_idx = i
else:
tot_n += i_n
if i_start < tot_start: tot_start = i_start
if i_stop > tot_stop: tot_stop = i_stop
if purge_idx >= 0:
# clean the old records out of the window (slide the window)
self.window = self.window[purge_idx+1:]
if tot_n > 0:
# recompute; stay within bounds
self.sleep_interval = (tot_stop - tot_start) / tot_n
if self.sleep_interval > self.max_sleep:
self.sleep_interval = self.max_sleep
if self.sleep_interval < self.min_sleep:
self.sleep_interval = self.min_sleep
def _fill_cache(self):
"""Internal method for grabbing as much data out of the file as is
available and caching it for future calls to nextline(). Returns
the number of lines just read.
"""
old_len = len(self.queue)
line = self.f.readline()
while line != "":
self.queue.append(line)
line = self.f.readline()
# how many did we just get?
num_read = len(self.queue) - old_len
if num_read > 0:
self.pos = self.f.tell()
now = time()
self._recompute_rate(num_read, self.last_read, now)
self.last_read = now
return num_read
def _dequeue(self):
"""Internal method; returns the first available line out of the
cache, if any."""
if len(self.queue) > 0:
line = self.queue[0]
self.queue = self.queue[1:]
return line
else:
return None
def _reset(self):
"""Internal method; reopen the internal file handle (probably
because the log file got rotated/truncated)."""
self.f.close()
self.f = open(self.path, "r")
self.pos = self.f.tell()
self.last_read = time()
def nextline(self):
"""Return the next line from the file. Blocks if there are no lines
immediately available."""
# see if we have any lines cached from the last file read
line = self._dequeue()
if line:
return line
# ok, we are out of cache; let's get some lines from the file
if self._fill_cache() > 0:
# got some
return self._dequeue()
# hmm, still no input available
while True:
sleep(self.sleep_interval)
if self._fill_cache() > 0:
return self._dequeue()
now = time()
if (now - self.last_read > self.max_sleep):
# maybe the log got rotated out from under us?
if stat(self.path)[ST_SIZE] < self.pos:
# file got truncated and/or re-created
self._reset()
if self._fill_cache() > 0:
return self._dequeue()
def close(self):
"""Close the tail monitor, discarding any remaining input."""
self.f.close()
self.f = None
self.queue = []
self.window = []
def __iter__(self):
"""Iterator interface, so you can do:
for line in filetail.Tail('log.txt'):
# do stuff
pass
"""
return self
def next(self):
"""Kick the iterator interface. Used under the covers to support:
for line in filetail.Tail('log.txt'):
# do stuff
pass
"""
return self.nextline()
Espero que te sea de ayuda. A mi me sirvió y me sigue funcionando :-)
Saludos
Esteban
El 29/03/2011, a las 01:33, Carlos Herrera Polo escribió:
Gracias a todos por sus aportes, estoy implementando un lector de logs
utilizando las ideas del link que me sugerio Angel
http://stackoverflow.com/questions/136168/get-last-n-lines-of-a-file-with-python-similar-to-tail/136368#136368
Muchas gracias a todos por sus aportes y ayuda.
El 28 de marzo de 2011 18:30, Chema Cortes <pych3m4 en gmail.com> escribió:
El día 27 de marzo de 2011 22:33, Carlos Herrera Polo
<carlos.herrerapolo en gmail.com> escribió:
> Alguien que me pueda sugerir algo por favor ?
Se puede hacer fácilmente, y sin recurrir a herramientas externas,
manteniendo el fichero abierto para lectura y leyendo de él cuando se
detecte que ha cambiado su tamaño. Algo parecido al código de este
mensaje:
http://mail.python.org/pipermail/python-es/2003-May/002091.html
_______________________________________________
Python-es mailing list
Python-es en python.org
http://mail.python.org/mailman/listinfo/python-es
FAQ: http://python-es-faq.wikidot.com/
El día 27 de marzo de 2011 22:33, Carlos Herrera Polo
<carlos.herrerapolo en gmail.com> escribió:
> Ilustres, quisiera tener su opinion al respecto..
>
> Administro ciertos servidores linux, y consolido la informacion de los
> logs de estos en un unico servidor con el servicio syslog-ng, los logs
> son archivos planos que se almacenan en formato ASCII, un log por cada
> servicio/servicio.
> Quisiera desarrollar un programa simple que me alerte por correo de
> ciertos eventos que ocurren en los logs.... Existe un programa llamado
> logsurfer, escrito en C que hace esto...
> Mi idea es hacer un logsurfer pero en python, pero no se como trabajar
> la lectura de estos archivos planos que cada uno debe tener varios
> megas de peso, y son escritos cada segundo, ustedes como lo harian ?
> Porque hacer fileread por cada uno y luego irme hasta las ultimas
> filas como que no seria eficiente...
>
> Alguien que me pueda sugerir algo por favor ?
>
> Muchas gracias a todos
>
> Saludos
>
> --
> Enviado desde mi dispositivo móvil
> _______________________________________________
> Python-es mailing list
> Python-es en python.org
> http://mail.python.org/mailman/listinfo/python-es
> FAQ: http://python-es-faq.wikidot.com/
>
Más información sobre la lista de distribución Python-es