1. #!/usr/bin/python
2.
3. from pyinotify import *
4. import os, os.path
5.
6. flags = IN_CLOSE_WRITE|IN_CREATE|IN_Q_OVERFLOW
7. dirs = {}
8. base = '/log/lighttpd/cache/images/icon/u241'
9. base = 'tmp'
10.
11. class UpdateParentDir(ProcessEvent):
12. def process_IN_CLOSE_WRITE(self, event):
13. print 'modify', event.pathname
14. mtime = os.path.getmtime(event.pathname)
15. p = event.path
16. while p.startswith(base):
17. m = os.path.getmtime(p)
18. if m < mtime:
19. print 'update', p
20. os.utime(p, (mtime,mtime))
21. elif m > mtime:
22. mtime = m
23. p = os.path.dirname(p)
24.
25. process_IN_MODIFY = process_IN_CLOSE_WRITE
26.
27. def process_IN_Q_OVERFLOW(self, event):
28. print 'over flow'
29. max_queued_events.value *= 2
30.
31. def process_default(self, event):
32. pass
33.
34. wm = WatchManager()
35. notifier = Notifier(wm, UpdateParentDir())
36. dirs.update(wm.add_watch(base, flags, rec=True, auto_add=True))
37.
38.
在已经有Hash Tree的时候，同步就比较简单了，不停地获取根目录的修改时间并顺着目录结构往下找即可。需要注意的是，在更新完文件后，需要设置修改时间为原文件的修改时间，目录也是，保证Hash Tree的一致性，否则没法同步。mirror.py的代码如下
1. #!/usr/bin/python
2.
3. import sys,time,re,urllib
4. import os,os.path
5. from os.path import exists, isdir, getmtime
6.
7. src = sys.argv[1]
8. dst = sys.argv[2]
9.
10. def local_mirror(src, dst):
11. if exists(dst) and mtime == getmtime(dst):
12. return
13. if not isdir(src):
14. print 'update:', dst
15. open(dst,'wb').write(open(src).read())
16. else:
17. if not exists(dst):
18. os.makedirs(dst)
19. for filename in os.listdir(src):
20. local_mirror(os.path.join(src,filename), os.path.join(dst,filename))
21. os.utime(dst, (mtime,mtime))
22.
23. def get_info(path):
24. f = urllib.urlopen(path)
25. mtime = f.headers.get('Last-Modified')
26. if mtime:
27. mtime = time.mktime(time.strptime(mtime, '%a, %d %b %Y %H:%M:%S %Z'))
28. content = f.read()
29. f.close()
30. return int(mtime), content
31.
32. p = re.compile(r'([\d.]+?) +([\w/]+)')
33.
34. def remote_mirror(src, dst):
35. mtime, content = get_info(src)
36. if exists(dst) and mtime == int(getmtime(dst)):
37. return
38. print 'update:', dst, src
39. if not src.endswith('/'):
40. open(dst,'wb').write(content)
41. else:
42. if not exists(dst):
43. os.makedirs(dst)
44. for mt,filename in p.findall(content):
45. mt = int(float(mt))
46. lpath = dst+filename
47. if not exists(lpath) or int(getmtime(lpath)) != mt:
48. remote_mirror(src+filename, lpath)
49. os.utime(dst, (mtime,mtime))
50.
51. if src.startswith('http://'):
52. mirror = remote_mirror
53. else:
54. mirror = local_mirror
55.
56. while True:
57. mirror(src, dst)
58. time.sleep(1)
如果源文件不在同一台机器上，可以通过NFS等共享过来。或者可以通过支持列目录的HTTP服务器来访问远程目录，mirror.py 已经支持这种访问方式。server.py 是用webpy做的一个简单的只是列目录的文件服务器。由于瓶颈在IO上，它的性能不是关键。server.py的代码如下：
1. #!/usr/bin/python
2.
3. import os,os.path
4. import web
5. import time
6.
7. root = 'tmp'
8.
9. HTTP_HEADER_TIME = '%a, %d %b %Y %H:%M:%S %Z'
10.
11. class FileServer:
12. def GET(self, path):
13. path = root + path
14. if not os.path.exists(path):
15. return 404
16. mtime = time.localtime(os.path.getmtime(path))
17. web.header('Last-Modified', time.strftime(HTTP_HEADER_TIME, mtime))
18. if os.path.isdir(path):
19. for file in os.listdir(path):
20. if file.startswith('.'): continue
21. p = os.path.join(path,file)
22. m = os.path.getmtime(p)
23. if os.path.isdir(p):
24. file += '/'
25. print m, file
26. else:
27. print open(path,'rb').read()
28.
29. urls = (
30. "(/.*)", "FileServer",
31. )
32.
33. if __name__ == '__main__':
34. web.run(urls, globals())
