Log Monitor
公司的系统很分散，monitor log很费劲。写了一个ruby程序，让所有的log集中在一个地方。试用以后，系统维护的工夫可以少很多，还可以在用户报告之前发现问题。效果不错。主要原理就是用net/ssh登录各台server，tail -f那些log，捕捉sysout（channel.on_data），然后加一些辅助信息（比如--orignal log）输出。
以前也用过chainsaw等工具，不方便的地方在于：只能monitor log4j；不能集成在一个output里。
这个ruby程序只有几十行，如果以后要加一些auto mail，垂直log等功能也很灵活。还有些说明与窍门都一一分享在code comments里。
### license: http://creativecommons.org/licenses/by-sa/2.5/deed.zh### by: caoweiyuan#gmail.comrequire 'net/ssh'require 'yaml'modules = YAML.load(File.open('modules.conf'))### sample config# engine-1:# username: admin# password: ******# logs:# /opt/apps/production/tomcat/logs/app.log:# name: # greps:# error.log: # - ERROR# info.log: # - INFOdef do_grep(name, line, greps)log = "[#{name}] - #{line}" # default output to screenputs log # grep patterns to sepecific filesfor log_file, patterns in greps pattern = (patterns.class == Array) ? patterns.join("|") : patterns File.open('log/' + log_file, 'a') {|f| f.puts log} if line =~ /#{pattern}/ # TODO: auto mail for errors # TODO: horizontal log for different products, and delete after each process over.endenddef do_tail( session, log_file, log_conf )session.open_channel do |channel| channel.on_data do |ch, data| data.each_line do |line| do_grep log_conf['name'], line, log_conf['greps'] end end channel.exec "tail -1f #{log_file}"endendfor key, value in modulest = Thread.new(key, value) do |host, host_conf| Net::SSH.start( host, host_conf['username'], host_conf['password'] ) do |session| puts "#{host} logged in." for log_file, log_conf in host_conf['logs'] puts "start watching #{log_file}" # open channels for multiple log files do_tail session, log_file, log_conf end # log session.loop # important: loop to keep reading channel outputs end # sessionend # threadend # host# TODO: implement a formal deamonloop do# deamonizesleep 1end
最后，还有一点要分享的，这种方式不支持windows openssh，如果log在windows机器上，要装cygwin。还有，vmware的windows装了cygwin也不行，sshd开不了，如果有高手知道原因和方案，请不吝赐教。
页:
[1]