A Close Shave

There is a definite difference between the i386 Partition Table on disk and the one the kernel knows about. On disk it might say your 2 Terabytes are still there, but you cannot mount them. And you cannot figure out why. The disks are there but any command issued will return "No such device"
Well, if you did not know the above difference, you might get your backup tapes and start updating your resume. If you did know, you would issue "partprobe" and everything would be fine.
- from one who had to find out...

Unix Design Flaw

This has been a niggling thing in the back of my mind for some time now:
Consider an installation with central authentication. The root-account is local to each machine, all other accounts get drawn from, say, LDAP. Now any root on any client can assume any identity that gets served by, say, LDAP and thus read that identities files on any fileserver. So you say, ok we don't give them root on the workstations but that takes effort and is hard to police. Also anyone can install a system in minutes. So you have to secure your DHCP. Current solution then is:
1. Workstations root-account barred.
1a. (Maybe some use of sudo. sudo itself has problems, too, though. )
2. LDAP Server Access Secured with krb5 and SSL.
3. DHCP secured against new workstation, or new workstations get "guest status" in a different VLAN.

...but that is a hard one to police. Looking for simplification.
Comments welcome.

Bash vs Perl in Administration

Ok, so I am new to Perl. I confess. Normally I'd say "I am a bash guy!" and sort of brush it off. But today I hit the limits. Task at hand: "Compare two files with lists of md5sums".
"diff" does not quit ecut the biscuit, as it is to unstructured. So I want to take the first md5sum from the second file and remove it from the original. In that way the remaining entries are the ones different in the second file.
So in bash this spells out:

while read line
do
        # isolate md5sum from line:
       md5=$(echo $line| awk '{print $1}')
       # Is this md5 in the second file ?
       if grep -q "$md5" RESTORE-sorted.txt
       then
                # If so, throw it out, we don't consider it anymore:
               grep -v "$md5" RESTORE-sorted.txt > RESTORE.mv
               mv RESTORE.mv RESTORE-sorted.txt
       fi

done 

Pretty short and sweet. However, it runs forever. on a 650M file. Something to do with the kernels handling of file-descriptors. I started it 6 hours ago and it has not even done half of the task. In fact, while it was running I was able to pick up the necessary Perl to accomplish the same, using arrays. (Perl is quite "intuitive", you can sort of "baby-talk" your way into it) The prog is not quite as short and sweet, but that is probably due to my newbieness. However it takes 10 seconds to run. Well, does illustrate a point, does it not...

#!/usr/bin/perl

# Read original file and checksums into array:
$orig_file="ORIG-sorted.txt";
open(ORIG, $orig_file) || die("Could not open file!");
while ()
{
 ($key,$value) = split(/ +/,$_);
 $orig_a{$key} = $value;
}
close(ORIG);

# Open the next file:
$restore_file="RESTORE-sorted.txt";
open(RESTORE, "<$restore_file") || die("Could not open file!");
while(<>)
{
       # Split the line in two parts...
       my($line) = $_;
       @record = split(/ +/,$line);
       ##...and delete line containing md5sum from original array:
       ##(the central task)
       delete $orig_a{"$record[0]"};

};
close(ORIG);

# print out formatted array:
foreach $key (keys %orig_a)
{
       print $key , " " , $orig_a{$key} ;
}

Re-activating Blog Activity

OK, Everyone must have a blog. If you don't blog, you don't live. That's why I decided to re-activate. Current activities are still Linux, now a bit more into the Debian world of things and lately VMware Infrastucture.