progress

I actually cleared off an approximately 10 X 14 inch area on my desk today.  My coworkers were laughing at me for being happy about this given how messy my cube still is, but since that area hadn’t seen the light of day in something like two years (it’s always been the spot I NEVER get to), it was actually quite the project.

I really feel like I had an unproductive week this week, probably because I spent more time than I have in years waiting on compiles, because actually, I got a lot of long needed work done or to the point where it can be scheduled, just nothing spectacular.

Advertisement

Thoughts to ponder

I found the following while cleaning my cube this morning:

• consistency and standardisation
• automation
• preprocessing of feedback
• simplification – KISS

Systems Administration is all about the above basic principles.Bishop, Craig.  “Applying SysAdmin Techniques to Security”.  SANS System Administration Networking and Security Conference, Best Practices in System Administration, Networking & Security, System Administartion, Web and Networking and Security Tracks: Technical Conference (Proceedings).  April 20-26, 1997.

(I picked this up in the free bin at the entrance to Powell’s Technical Books and pretty much dumped on my desk behind the CPU until I have a chace to look it over….)

A Failure and a Pass

Well, I failed the MS exam – then again, 560 something isn’t bad when you didn’t study and don’t use the OS regularly.  (700 is passing.)  Actually, I found the exam in many ways easier than I would have expected.  Kind of funny to say that when I failed, but I expected the questions to be more technically or intellectually rigorous, but they were more about details of fairly basic stuff.

On the up side, I got my results from the LPI 101 exam I took at OSCON, which I did pass even though I don’t use Linux much either (but I do use the GNU and Unix commands that are also tested.)  Ironically the section I did worst on was X, which I do use, though I try very hard to not administer X Windows whenever possible.

So I guess I need to study to retake the MS exam and study for LPI 102 now.

Windows Administration Training

I am by no means a Windows Administrator and I’ve even been known to admit to being a Unix bigot; but really I agree with the accumulated wisdom of ASR that all operating systems suck, some just suck more than others (see also FAQs 3.1 & 3.3.)  And I don’t believe that a good system administrator can be judged by the operating systems they administer (the reasons they administer those operating systems maybe, but not the operating systems.)  At any rate, it’s useful to have a basic understanding of other OSes and the last two years my training plan has included a Windows Administration class at the local community college.  I’ve been thinking that I should take the associated Microsoft Certification exams, but I’ve never really gotten around to it.  However, for some reason that escapes me now, I signed up to take 70-291 Windows Server 2003 Network Infrastructure this afternoon.  (I remember the part about the VUE second chance promotion, but not why I thought I’d actually have time to prepare.)  So, I spent a good part of the weekend trying to force myself to read a study guide.  Given that I only got through about a third of it, it’s a good thing that a lot of the curriculum for this exam is about TCP/IP, subnetting, and protocols – all things I already know – though I’m not making any predictions on success or failure.  (But hey, if I fail I get a free retake and maybe I’ll actually do a through review for that.)  The thing I’m always amused at when reading Windows adminstration training material is how they seem to use the following formula repeatedly in almost every book:
Topic is really difficult.

4 to 8 step procedure with no decision points required (and often with illustrations)

See you did that really difficult thing, you are so smart.
I don’t disagree that Windows administration is hard and requires skilled administrators, but the hard stuff can’t be laid out in a 4-8 step unchanging procedure.  And I think all this does is build fake self-esteem on the part of the poor innocents who don’t know enough to know how little they know yet (and provide endless amusement for the rest of us when we hear them bragging about it…. maybe it’s a huge elaborate practical joke?)

More on make

I think I’ve found a bug in Solaris make.  I’ve spent lots of time this week trying to fix the dependancy problem in my makefile for package building that was causing my rule to convert a tarball to a directory to ALWAYS run. Turns out the problem only exists in Solaris make (GNU Make on linux or Solaris, and AIX make work fine). I know lots of people say you shouldn’t use directories as make targets, but there are situations where you want to do something if a file has been added to a directory or do something with a directory if a file has changed.  In particular, if I get a new tarball, I want to untar it and recompile, if I’ve done that I want to build a new packaging directory, if I’ve updated the packaging directory, I want to build a new package; Solaris packages themselves are directories by default and I like to convert them to filestream format for ease of use; and so forth.

The good news is that I’ve learned a lot about make while trying to figure this out.  At this rate the pages in the “make” section are going to start falling out of my copy of UNIX in a Nutshell along with the awk and sed sections and the page on “test” – not that it’s not a miracle that it’s even still held together at all given how much I use it.  And I’ve finally actually read most of Managing Projects With Make.

I tried to report the make problem to Sun, since we have support contracts, but someone forgot to put me on the list of technical contacts when we renewed our contracts.  Admittedly, most of the other SA’s here just have the operators call Sun for them so I guess it’s a reasonable mistake – but one of the things I like about working here is that we have good support and if I can’t use it, it’s useless to me.  (Of course, the drawback of having support is that at some point you are supposed to use support instead of tinkering with whatever the problem is yourself even when it’s fascinating.)  Hopefully the contract wrinkle will be worked out sometime next week.  It’s not like I can’t just use GNU make, but heck we pay for this OS, when it’s broken, it should get fixed.

sed and make

It’s funny how you can go years without using tools and then start using them all the time.  I haven’t done anything with sed other than the occasional sed ’s/foo/bar/’ in a pipeline since I got comfortable with awk (something like 1994 or 1995), but all of a sudden I find I’m using it daily, and not just the s command.

Part of this has to do with my new obsession with make.  I’m lazy and not much of a programmer, so while I usually use make to compile, if I have to write a Makefile, I usually use:touch Makefile
and just use the default rules.  However, I recently came up with the idea of writing a makefile that could be used to build a Solaris package from a software package that would normally be built using “tar xvzf X.tar.gz; cd X; ./configure configure options; make install”  Actually, my original idea was to build a Makefile just for building AMANDA packages, since as a lazy person I usually get my software from Sunfreeware.com but I quickly decided it would be more useful and interesting to generalize it.  Then yesterday, I sat down to setup the final patchlists for the 5 servers I’d scheduled for patching this morning and decided that make would be a good way to further automate my process (I’ve been using a script, but then I run into trouble when I only want to run part of the process for some reason – like creating my patch list for  prod and test today but only backing up configuration files that will be changed for test today.)

For some reason, I find that sed seems much cleaner than awk in a Makefile.  (This might be because, on Solaris at least, you seem to need to write your sed script as sed -e cmd1 -e cmd2 …)  I suspect that sed is much more efficient than awk or nawk, but I’ve never profiled them (hmmm, there’s a good activity for a dull day…)

However, make is not the only source of my renewed love affair with sed.  I think it actually started in my personal life.  I’m taking a Spanish class and each class we learn one or more tricks for recognizing (and converting) English words that are similar or identical to Spanish words.  I’ve started writing my notes about this in sed (I actually wrote a script to run against the system dictionary as well, but I quickly converted it to nawk so I could do nicer formatting without pulling in nroff.)

Why am I here?

I’d love to say something brilliant about why I’m doing this, but I don’t have a brilliant reason , only a niggling feeling that it might be useful.

A few months ago some of us from PASA (Portland Area System Administrators) had a short discussion about system administrator blogs, and all of our first thoughts fell along the lines of “why would you want to read that.”  But as I’ve thought about it some more, I’ve started to wonder if blogging might be an ideal way to represent and organize some of the information that I either don’t currently keep well at all or that I lose track of and spend a lot of time looking for later.  My first boss as an SA, Tom from FASTIXX, required that we all keep a notebook with notes in it day by day.  I’ve adopted that as a work habit (although I’ve upgraded to bound books with int
eresting covers which sometimes helps me remember which notebook I wrote in) and think of it as sort of a lab notebook (although rarely used so formally and per
fectly) where I write down notes and ideas about the work I’m doing, what I’ve tried, what worked, what didn’t.   This works really well for me as a reference, except that when I need to go back to something, I often find myself paging through 2 or 3 books of notes looking for it, because the best I can recall was it was sometime in the spring of `99 or something like that.  I do keep a text file of stuff I expect to have to repeat or refer to again, but that usually contains the solution, not the steps I took to get there or what didn’t quite do what I wanted then but is close to what I want to do today.

I think too this might be a better way to track what I’ve actually accomplished, which might result in my doing a better job of both sharing the results and reporting how I’m spending my time at work to my manager.  For instance, sometimes I realize later that something that started out as a one-line hack has by small incremental improvements become an extremely useful general sysadmin tool, I guess I hope this helps me remember to share that realization with my boss and share the tool beyond my work group.