Having several processes share a logfile

[ibm1620]

I'm designing a multiforking server and I'd like to have all the processes write to the same logfile (correctly interleaved, of course). Is there a module recommended for that?

[pemungkah]

This may be of interest; it shows you how to set up an appender in Log::Log4Perl so that you get atomic writes.

[davido]

The POD for flock demonstrates how to obtain an advisory lock, and then seek to the EOF before appending. The example is pretty simple. Look at over and I'm sure it will make sense.

[ibm1620]

This is what I ended up doing - easy enough for development. Thanks! Chap

[flexvault]

Hello ibm1620,

As was already mentioned, I use the *nix 'syslog' in production. But since you're in development mode at the present time, I suggest you use a unique log file for each 'fork'ed child.

     use Sys::Syslog qw(:DEFAULT setlogsock)
     our ( $LOG, $Debug, $start );
     $Debug = 4;           ## I use 0 for production, and 1..9 for tes
+ting
     our $NAME = "$prog-$Debug";   ## $prog should be a name to help y
+ou
     setlogsock('unix');
     openlog(" $$ Srv", 'ndelay', 'localn');    ## 'n' is defined by s
+ystem admin
     syslog('info', "############# Starting #############");
     . . .
     if ( $Debug == 0 )     ## This is production 
     {   open ($LOG, ">","/dev/null") || Die_RTN("$NAME: Not open-8 /d
+ev/null");
     }
     else                   ## This will log for levels 1..9 
     {   open ($LOG, ">","./logs/Child$$") || Die_RTN("$NAME: Not open
+-9 Child$$");
         my $time = time; $start = 0;
         print $LOG "Info: |$time|$start|\n";
     }
     $start++;
     if ( $Debug >= 4 )     ## This is for level 4 or greater debug in
+formation 
     {   my $ctime = time;
         print $LOG "Info: |$ctime|$start|\n";
     }
     . . .
     closelog(); close $LOG;

This allows you to see what logs are generated for the production environment, and also allows you to have very detailed debugging information during development. Once code is working correctly, you can make the test for a higher number and eliminate unneeded data from the child logs.

I open the 'syslog' before forking the children, and then open the debugging logs in the children. The '$$' is the unique process ID of the child.

Hope this helps and that I typed everything correctly :-)

Good Luck...Ed

[Anonymous Monk]

How about using the syslog facility on *nix? A few Perl interfaces to it can be found on CPAN.

[nagalenoj]

I've used Log::Log4perl and it worked fine for the same need.