Contrary to an earlier suggestion, I'd read "file 1" first, and store each "user*" string as a hash key -- there will be less data overall to keep in memory.

Then, reading "file 2", only print a line if its first token does not exist as a hash key. These are the records in file 2 that do not exist in file 1.

Put that way, the script can be very short, especially if you just give the file names as command line args:

use strict;

die "Usage:  $0  file.1  file.2\n"
    unless ( @ARGV == 2 );    # test for number of args

my %known;

open( F, "<", $ARGV[0] ) or die "$ARGV[0]: $!";
while (<F>) {
    chomp;
    $known{$_} = undef;  # don't need a value, just the key
}
close F;

open( F, "<", $ARGV[1] ) or die "$ARGV[1]: $!";
while (<F>) {
    my $key = ( split ' ', $_ )[0];  # get first "word"
    print unless exists( $known{$key} );
}
close F;
[download]

That will print the targeted records to STDOUT, which you can redirect to a file via the command line:

little_perl_script.pl  file.1  file.2  > target.set
[download]

(update: just noticed that this is identical to TedPride's suggestion, ignoring a few trivial, irrelevant differences. sorry about the redundancy)