perl question needing to get value from file

Question

I am writing a perl script to automate certain processes on our Oracle databases. One portion of the script involves putting the script to sleep until the alert log shows the database open.

I intend to do a sleep until (logevalmodule). The logevalmodule will be a sub procedure that opens and monitors the alert log using tail -f returning a true when the word OPEN appears in the log.

Can anyone help me with possible solutions. I have two hurdles to make this work. The path to the alert log will not be in the same place for all the databases on the Oracle server, for example 'apps13/oracle/admin/db1/bdump' and 'apps14/oracle/admin/db2/bdump'. The other hurdle is to format the tail -f into the script.

Thanks.

Answer 1

In terms of building a Perl module to monitor stuff appearing in Log files I would strongly recommend using Perl Object Environment (POE). I've built a bunch of log monitors using it in the past and it's pretty simple and quite effective. There's a good example showing some tail monitoring\processing of a web server log on the POE Cookbook Site here.

As far as the other part of the problem all I can suggest is set up a bunch of POE Wheels targeting each log you need to monitor, populated in some way that suits you. Personally I'd just read the desired target locations in from a file.

Answer 2

Regarding log location:

1) If you have the right privileges on Oracle, you can get the location of the bdump directory by querying v$parameter for "BACKGROUND_ DUMP_DEST".

2) bdump will usually be in $ORACLE_BASE/admin/$ORACLE_SID/bdump The existence of these environment variables is mandatory for user "oracle" or they can be parsed from /etc/oratab

Answer 3

How about something like:

sub delay_until_open
  {
  my( @log_files) = @_;
  open( ALL_LOGS, 'tail -0f ' . join( ' ', @log_files) . ' |') || die 'oops, it broke';
  while ( <ALL_LOGS>)
    {
    if ( /OPEN/)
      {
      last;
      }
    }
  close ALL_LOGS || die 'oops, it broke';
  }

Obviously, the error handling could use some work, but this should scan all of the log files, then end the function when one of them contains (newly appends) OPEN. You won't need a sleep + check loop around this.

In addition, since "tail -f" on multiple files displays the file name when a chunk comes out from one of the files, you could capture the file name in a regex, and return the name of the log file that just announced an "OPEN". Another caveat, doing this as a single process, instead of a 2 step pipeline, means that you may miss some events while you are servicing earlier events.