bmalkow@malkowscy.net

Backup repozytorium

Znalazłem na sieci śliczny skrypcik do archiwizowania repozytorium svn:

#!/usr/bin/python

# script to backup SVN repos at a given time from the last revision number
# Greg Loscombe

import os
import sys

debug = 0


# build a list of projects and what the last revision backup

basedir = "/home/subversion/basedir"
basebackup = "/var/log/svnbackup"
revlogfile = "/var/log/svnbackup/last_rev" 
projects = {}
something_dumped = 0

def get_youngest(proj):
   youngest_cmd = "/usr/bin/svnlook youngest " + basedir + proj
   if ( debug == 1):
      print "command:", youngest_cmd
   cmdout = os.popen( youngest_cmd )
   rev_youngest = cmdout.read()
   rev_youngest = int( rev_youngest.strip() )
   cmdout.close()
   return rev_youngest

def dump_incremental(base, proj, rev_from, rev_youngest):
   location = base + proj + ".dump." + str( rev_from ) + "-" + str(rev_youngest )
   if ( debug == 1):
      print "Location dumped too", location
   incremental = "/usr/bin/svnadmin dump --incremental " + basedir + this_proj + " -r " + str( rev_from ) + ":" + str( rev_youngest ) + " > "+ location
   cmdout = os.popen( incremental )
   cmdout.close()
   if ( debug == 1):
      print "Command incremental dump:-", incremental

def update_log():
   revfile = file( revlogfile, "w")
   for this_proj in projects.iterkeys():
      revfile.write( this_proj + " = " + str(projects[this_proj]) + "\n")
      if ( debug == 1):
         print "New log:-", this_proj + " = " + str(projects[this_proj]) + "\n"
      print this_proj + " = " + str( projects[this_proj] )
   revfile.close()

def update_error_log(message):
   errorfile = file( "/var/log/svnbackup/error_log", "a")
   errorfile.write( message )
   #errorfile.write( proj + ":" + str( updated_rev ) + ":" + message )
   #add more stuff here soon
   errorfile.close()

#def update_access_log(message):
#this will print things like last run etc

# main part now, lets get all the old revision numbers we backed up

revfile = open( revlogfile )

while 1:
   line = revfile.readline()
   if line == "":
      break
   if ( ( line.startswith("[") or line.startswith("#") ) is False ):
      proj, last_rev = line.split(' = ')
      last_rev = last_rev.strip()
      projects[proj] = int(last_rev)

revfile.close()

# lets get the new youngest and compair, if its newer, then do a incremental dump
# if its older than the last log, we have a problem, log it
# if its the same nothing happens
# if its never been dumped, time for a full dump
for this_proj in projects.iterkeys():
   rev_youngest = int( get_youngest(this_proj) )
   rev_old = int( projects[this_proj] )
   if ( debug == 1 ):
      print "this project", this_proj, "old revision", rev_old, "youngest", rev_youngest

   if ( rev_old > rev_youngest ):
      errmessage = "logged backed up revision " + str(rev_old) + " > youngest current revision " + str(rev_youngest) + "\n"
      #update_error_log(this_proj, rev_youngest, errmessage)
      update_error_log(errmessage)

   elif ( rev_old == 0 ):
      print "First time dump, do it all for", this_proj, "upto", rev_youngest
      dump_incremental(basebackup, this_proj, 0, rev_youngest)
      something_dumped = something_dumped + 1
      projects[this_proj] = int(rev_youngest)

   elif ( rev_old < rev_youngest ):
      print "Update for project", this_proj, "upto", rev_youngest
      rev_from = rev_old + 1
      dump_incremental(basebackup, this_proj, rev_from, rev_youngest)
      something_dumped = something_dumped + 1
      projects[this_proj] = int(rev_youngest)

   elif ( rev_old == rev_youngest ):
      print "Revisions have not changed for project", this_proj

# if something changed then its time to update the last logged file
if ( something_dumped > 0 ):
   print something_dumped, "repos have changed since last dump"
   update_log()
else:
   print "Nothing has changed, nothing done"

# the end

W pliku /var/log/svnbackup/last_rev należy umieścić nazwę repozytorium i numer ostatniej zarchiwizowanej wersji, np.:

/jaxmpp = 0
/proplan = 0

Całość dodać można sobie do crona, i już mamy eleganckie, przyrostowe backupy repozytorium.