Not to be taken personally, my petit prince; this ramble is general, not targetted.
Quoting The Little Prince ([EMAIL PROTECTED]): > On Thu, 27 Feb 2003, Chuck Yerkes wrote: > > Can you have multiple deliverers and poppers over NFS? > > Sure, write your own locking, rewrite mail.local and qpopper chunks, > > deal with the fact that NFS performance is likely 5-10% the > > performance of RAID for several times the cost. > > > > Never quit, because only you can run it. > > just another reason to justify a raise. :-) > > my boss always used to tell me..man, if you ever got hit by a bus, we'd be > dead. kinda makes you feel like one of those metal briefcases people > handcuff to their wrists. Yeah, I had guys who'd wack something together (and used to be one of those). But the "hit by a bus" can also be pronounced "take a vacation". If you feel cool cause you can't take a week off without being paged, then your business is weaker because of you and you are failing. If you can't leave for 1 or 6 months, then you haven't build a sustainable environment. It's about professionalism in many ways. unfo, several bad products get put in place because cowboy system admins make businesses wary and put them in danger. Yes, there is certainly the "clues for the clueless" - being innovative and bringing in good things and strengthening your company. But it's sometimes hard to tell that from the feeling you get because you've got a co-dependant company that you've rigged to nee *you* just cause you do things differently. I'll offer two examples. There are many machines running different things. Admins keep windows open to all and watch top and watch for them to crash. When there is a problem, they leap into action. For problem after problem. A System Admin Goofus and Galant, if you will: 1) You've setup Net-SNMP to fully monitor and even be reactive on several of your systems. BigBrother/Sister or Nocol let you and your cohorts be paged when certain traps are emitted. Scripts will take and parse data and look for combinations of things to trip alarms (all of the WAN is down, but so are your WAN routers and UPSs - maybe the WAN is fine and the problem is your routers). The NetOps folks can monitor your machines with HP OpenView; your new SAGE Level 2 SA can take a couple of the SNMP-triggered scripts and modify them to make them a little smarter. While you were skiing, with no cell coverage, a machine ate it. SCSI board went bad. Alerts went off and a cold spare was brought in. Worst case they had to kickstart/jumpstart a new boot disk that made the new machine into the dead server. You learn about it on Monday. You go about getting a more CPUs for your database machine cause the load on it has been increasing by about 10% a quarter and you need to get some budget for it. 2) You've written something that listens on a port and can report machine and application statuses. You've got scripts watching machines, running mail, managing LDAP. It's great - they had NOTHING before and were all just reacting to fires before you came in. All the scripts are written in Python. Why Python? Because it's the only language you'll deign to use. Perl is crap. Java is a pig. C is an old joke. Oh, nobody else knows Python and you are the GoTo guy for all problems with this system. You get raises; you get praise; the CTO needs you. You take the afternoon to hit Fryes and pickup a new motherboard cause you need to make that server a bit faster and want to get it in my next week. You impress your friends with how important you are - this company DEPENDS on you - as you get paged through the evening; coming back they ask what's up? "I just saved their asses... Again!" you crow. One is good, one is bad; both feel about the same. Be careful, cause you're never the bad guy. If a reasonable business truly really needs you, you're not doing your job well enough. You should be valuable but not indispensable. And remember, a boss once told me, NOBODY is indispensable. It's worthy of periodic self examination to discern where your work is on that spectrum.
