Theoretical vs Practical Knowledge

Wed Dec 12 23:28:55 PST 2001

Well, obviously you mount the mailq noatime and
you set syslog to stop fsyncing (many  modern syslogs
allow this, sun's doesn't).

I have a boss now who is infamous to the people to came after
him for "describe all the options to ls." It's an icebreaker.
When a "sr system admin" manages to list 4 and stop then you
know s/he's faking it.  The sr. people know that it's pretty
large.

And yeah, coming from the era of NSF funded connections,
when I get nervous I transpose those at times.  Hell, my
film background has caused me to talk about SMPTE protocols :)

Quoting J C Lawrence (claw at kanga.nu):
> On Wed, 12 Dec 2001 17:50:02 -0800 (PST) 
> Roy S Rapoport <rsr at inorganic.org> wrote:
...
> That's the entire concentration of my interviewing technique for
> SysAdms.  I tell them up front that I'm going to set scenarios for
> them, and then ask them to evaluate and resolve them by telling me
> what they'd do with me telling them the results.  I'm also careful
> to tell them that I don't expect instant right answers, and that in
> fact I do expect them to NOT resolve some or several of them as what
> I'm interested in is their problem solving approach and technique.
> I then just make up (or remember) various things and throw them at
> them.
> 
>   Okay, you've just inherited a dept's mail server which is acting
>   as the smarthost for the corporate mailing lists.  You're told
>   that historically it has been running just fine, but that
>   currently mail is NOT being delivered in any sort of timely
>   fashion.  Please get mail being delivered quickly.
> 
>     Solution to that one BTW is /var/log being on the same spindle
>     as /var/spool/<MTA> and syslog being configured to fsync() after
>     every write for the MTA logs, thus IO throttling the box, with
>     the change in behaviour due to a recent corporate mktg campaign
>     increasing mail loads enough to hit saturation.  And yes, that
>     happened to me and took me most of a day to figure out...
> 
> etc etc etc
> 
> It makes interviews long unfortunately (Linda Bissum and Jim Dennis
> can attest to that), especially for the good people, but it also
> weeds out the pretenders quickly and thoroughly (eg those who are
> not sure how to spell "NFS").