raid_watcher.rb

#!/custom/ree/bin/ruby
require 'net/smtp'

class RaidWatcher
  def self.send_alert msg
    begin
      domain = $config.email.domain,
      host   = $config.email.server,
      port   = $config.email.port,
      from   = "raid@#{$config.email.domain}"
      to     = "someone_who_can_fix_it@somewhere.com"
   
      content = '' 
      content << "From: #{from}\r\n"
      content << "Subject: #{`hostname -s`} RAID STATUS CHANGE!\r\n"
      content << msg

      Net::SMTP.start(host, port, domain) do |smtp|
        smtp.send_message(content, from, to)
      end
    rescue => e
      # this is bad.
    end
  end

  def self.check_status
    IO.popen('/usr/StorMan/arcconf getconfig 1 AL', 'w+') do |io|
      io.close_write
      begin
        while line = io.readline
          if line =~ /Controller\sStatus\s*:\s(.*)/
            send_alert "RAID controller status not optimal! Status changed to: #{$1}" if $1 != 'Optimal'
          elsif line =~ /Logical devices\/Failed\/Degraded\s*:\s(.*)/
            send_alert "RAID controller failed/degraded! Logical device status changed to: #{$1}" if $1 != '1/0/0'
          elsif line =~ /Status\s*:\s(.*)/
            send_alert "RAID bbu status not optimal! BBU status changed to: #{$1}" if ($1 != 'Optimal' && $1 != 'Charging')
          elsif line =~ /Capacity remaining\s*:\s([0-9]+)\sp/
            send_alert "RAID bbu power DRAINING!!! BBU only has: #{$1}% power left." if $1.to_i <= 85
          elsif line =~ /Status of logical device\s*:\s(.*)/
            send_alert "RAID logical device status not optimal! Logical device status changed to: #{$1}" if $1 != 'Optimal'
          elsif line =~ /Failed stripes\s*:\s(.*)/
            send_alert "RAID device has FAILED STRIPES!" if $1 != 'No'
          elsif line =~ /State\s*:\s(.*)/
            send_alert "RAID physical disk is NOT online!! Physical disk state changed to: #{$1}" if $1 != 'Online'
          end
        end
      rescue EOFError
        # done
      end
    end
  end
end

RaidWatcher.check_status