考虑到我手上的服务器逐渐的增多,有时候需要大规模的部署同一个文件,例如因为方便使用systemtap这个工具定位问题,需要把手上几百台服务器同时安装kernel-debuginfo这个包,原有的方式采用一个源服务器,采用rsync或者scp之类的文件传输方式只能做到一个点往下分发这个文件,这个时候下发的速度就会比较的慢,基于以上原因,我写了一个基于bt协议传输文件的小工具,实际测试,传输到10个机房,70多台机器传输一个240M的这个内核文件,到所有的机器,源采用限速2m/s的上传速度,测试的结果大概只要140s,就可以全部传输完毕,这个效率是非常之高,如果不限速的情况下速度会更快,下面把这个程序开源出来。
#!/usr/bin/env python import libtorrent as lt import sys import os import time from optparse import OptionParser import socket import struct import fcntl def get_interface_ip(ifname): s = socket.socket(socket.AF_INET, socket.SOCK_DGRAM) return socket.inet_ntoa(fcntl.ioctl(s.fileno(), 0x8915, struct.pack('256s', ifname[:15]))[20:24]) def ip2long(ip): return reduce(lambda a,b:(a<<8)+b,[int(i) for i in ip.split('.')]) def get_wan_ip_address(): interfaces = set(['eth0', 'eth1', 'eth2', 'eth3', 'em1', 'em2', 'em3', 'em4']) ip = '' for i in interfaces: try: ip = get_interface_ip(i) if (ip2long(ip) < ip2long('10.0.0.0') or ip2long(ip) > ip2long('10.255.255.255')) \ and (ip2long(ip) < ip2long('172.16.0.0') or ip2long(ip) > ip2long('172.33.255.255')) \ and (ip2long(ip) < ip2long('192.168.0.0') or ip2long(ip) > ip2long('192.168.255.255')): return ip except: pass return ip def make_torrent(path, save): fs = lt.file_storage() lt.add_files(fs, path) if fs.num_files() == 0: print 'no files added' sys.exit(1) input = os.path.abspath(path) basename = os.path.basename(path) t = lt.create_torrent(fs, 0, 4 * 1024 * 1024) t.add_tracker("http://10.0.1.5:8760/announce") t.set_creator('libtorrent %s' % lt.version) lt.set_piece_hashes(t, os.path.split(input)[0], lambda x: sys.stderr.write('.')) sys.stderr.write('\n') save = os.path.dirname(input) save = "%s/%s.torrent" % (save, basename) f=open(save, "wb") f.write(lt.bencode(t.generate())) f.close() print "the bt torrent file is store at %s" % save def dl_status(handle): while not (handle.is_seed()): s = handle.status() state_str = ['queued', 'checking', 'downloading metadata', \ 'downloading', 'finished', 'seeding', 'allocating', 'checking fastresume'] print '\ractive_time: %d, %.2f%% complete (down: %.1f kb/s up: %.1f kB/s peers: %d, seeds: %d) %s' % \ (s.active_time, s.progress * 100, s.download_rate / 1000, s.upload_rate / 1000, \ s.num_peers, s.num_seeds, state_str[s.state]), sys.stdout.flush() time.sleep(1) def seed_status(handle, seedtime=100): seedtime = int(seedtime) if seedtime < 100: seedtime = 100 while seedtime > 0: seedtime -= 1 s = handle.status() state_str = ['queued', 'checking', 'downloading metadata', \ 'downloading', 'finished', 'seeding', 'allocating', 'checking fastresume'] print '\rseed_time: %d, %.2f%% complete (down: %.1f kb/s up: %.1f kB/s peers: %d, seeds: %d) %s' % \ (s.active_time, s.progress * 100, s.download_rate / 1000, s.upload_rate / 1000, \ s.num_peers, s.num_seeds, state_str[s.state]), sys.stdout.flush() time.sleep(1) def remove_torrents(torrent, session): session.remove_torrent(torrent) def read_alerts(session): alert = session.pop_alert() while alert: #print alert, alert.message() alert = session.pop_alert() def download(torrent, path, upload_rate_limit=0, seedtime=100): try: session = lt.session() session.set_alert_queue_size_limit(1024 * 1024) sts = lt.session_settings() sts.ssl_listen = False sts.user_agent = "Thunder deploy system" sts.tracker_completion_timeout = 5 sts.tracker_receive_timeout = 5 sts.stop_tracker_timeout = 5 sts.active_downloads = -1 sts.active_seeds = -1 sts.active_limit = -1 sts.auto_scrape_min_interval = 5 sts.udp_tracker_token_expiry = 120 sts.min_announce_interval = 1 sts.inactivity_timeout = 60 sts.connection_speed = 10 sts.allow_multiple_connections_per_ip = True sts.max_out_request_queue = 128 sts.request_queue_size = 3 sts.use_read_cache = False session.set_settings(sts) session.set_alert_mask(lt.alert.category_t.tracker_notification | lt.alert.category_t.status_notification) session.set_alert_mask(lt.alert.category_t.status_notification) ipaddr = get_wan_ip_address() #print ipaddr if ipaddr == "": session.listen_on(6881, 6881) else: session.listen_on(6881, 6881, ipaddr) limit = int(upload_rate_limit) if limit>=100: session.set_upload_rate_limit(limit*1024) session.set_local_upload_rate_limit(limit*1024) print session.upload_rate_limit() torrent_info = lt.torrent_info(torrent) add_params = { 'save_path': path, 'storage_mode': lt.storage_mode_t.storage_mode_sparse, 'paused': False, 'auto_managed': True, 'ti': torrent_info, } handle = session.add_torrent(add_params) read_alerts(session) st = time.time() dl_status(handle) et = time.time() - st print '\nall file download in %.2f\nstart to seeding\n' % et sys.stdout.write('\n') handle.super_seeding() seed_status(handle, seedtime) remove_torrents(handle, session) assert len(session.get_torrents()) == 0 finally: print 'download finished' if __name__ == '__main__': usage = "usage: %prog [options] \n \ %prog -d -f-s \n \ or \n \ %prog -m -p -s \n" parser = OptionParser(usage=usage) parser.add_option("-d", "--download", dest="download", help="start to download file", action="store_false", default=True) parser.add_option("-f", "--file", dest="file", help="torrent file") parser.add_option("-u", "--upload", dest="upload", help="set upload rate limit, default is not limit", default=0) parser.add_option("-t", "--time", dest="time", help="set seed time, default is 100s", default=100) parser.add_option("-p", "--path", dest="path", help="to make torrent with this path") parser.add_option("-m", "--make", dest="make", help="make torrent", action="store_false", default=True) parser.add_option("-s", "--save", dest="save", help="file save path, default is store to ./", default="./") (options, args) = parser.parse_args() #download(sys.argv[1]) if len(sys.argv) != 6 and len(sys.argv) != 4 and len(sys.argv) != 8 and len(sys.argv) != 10: parser.print_help() sys.exit() if options.download == False and options.file !="": download(options.file, options.save, options.upload, options.time) elif options.make == False and options.path != "": make_torrent(options.path, options.save)
准备环境:
需要在所有的os上面安装一个libtorrent的库,下载地址:
http://code.google.com/p/libtorrent/downloads/list
记得编译的时候带上./configure �Cenable-python-binding,然后mak,make install,进入binding目录,make,make install就
可以运行这个小的工具
当然大规模部署不可能采用每一台都去编译安装的方式,只要把编译出来的libtorrent.so libtorrent-rasterbar.so.7的文件跟bt.py这个文件放到同一个目录,另外写一个shell脚本
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$lib
python bt.py -d -f <种子文件> -s <文件保存路径> -t <做种时间> -u <限制上传速度>
使用方法:
首先在源服务器上面生成种子文件
发布文件
在源服务器上面,执行
其中做种时间默认设置的是100s,上传速度默认不限制,限制的速度单位是KB
下面的机器,直接可以
只要有一台机器完成了,就自动作为种子,在下载的过程中也会上传,任何一台机器都可以作为源服务器,当然了这里面还有中心的tracker服务器,脚本当中,我搭建了一个tracker源服务器,放到10.0.1.5端口是8760上面,当然大家也可以采用opentracker这个软件自己搭建一个tracker服务器,修改其中的源代码对应部分,另外考虑到发布都是私有文件,代码当作已经禁止了dht,如果还想更安全,就自己搭建一个私有的tracker server,具体搭建方法就使用一下搜索引擎,查找一下搭建的方法!
目前基本做到可以使用,后续考虑更简单一点,采用磁力链接的方式,这样就可以做到不用每台都要拷贝一个种子文件,采用一个单独的命令行就可以发布整个文件