it will make it to the end of the day, or do we re*oot it now, in the middle of the afternoon7 /s we ealuated the situation, we noted that the pro*lem likely would not cause any data loss, and we took into account that this was one of the *usiest days of the month! 1e decided to gam*le and leae the serer up until the end of the *usiness day, and it coasted along fine! 2ecisions like these are tough, *ecause you o*iously don"t want the serer to come crashing down if you can help it, *ut we decided that the minimum user impact would *e to leae the serer up! It was a decision *etween 9C minutes of guaranteed downtime for a re*oot, or 9C minutes of possible downtime if the serer crashed later and had to *e re*ooted! 1e did, howeer, take steps to notify all affected groups a*out the situation, which *rings me to my ne4t point! o
eep
the
users
informed
In my e4perience, users hae an easier time accepting a serice interruption if they know it"s coming! In the preceding e4ample, my group sent out a company-wide e-mail message adising the users that the serer was haing pro*lems and might crash, and telling the users to sae often! 1e got positie feed*ack a*out this approachK een though the users weren"t happy that the serer might die in the middle of the day, they appreciated the warning! ?ad the machine crashed, they would hae lost a lot less work than if they hadn"t had any
warning!
End users do like to know the *asic reasons for downtime, *ut it"s also a good practice to keep technical jargon out of end user communications! 1hile you may think it"s neato to e4plain the intricacies of E(E%)I. in an e-mail message informing users that the E4change serer will *e down for the weekend, the fact is that most people just don"t care! eep your communications straightforward and to the point, and your message will come across loud and clear! If an end user really is interested in what you"re doing with the E4change *o4, he or she will reply and ask for more information! or eeryone else, a simple note saying )he E4change
serer
will
*e
down
this
weekend
for
data*ase
maintenance will suffice! o
eep
the
help
desk
informed
1hen I worked on an enterprise help desk, it always seemed as if we were the last to know a*out ital information that would affect the users! ?alf the time, we gathered this information from the users themseles as they called,
which was terri*ly frustrating and
em*arrassing! .ike eery other tech support eteran I know, I swore up and down that once I got to the *ack of the shop, I"d make help desk communications a priority so that my first-leel support would neer *e without the proper information! ?ae I made good on this
promise7
Mostly,
although
I"m
still
not
as
communicatie to our help desk as I"d like to *e! It"s a good goal to hae, thoughK the *etter informed your first-leel support people are, the *etter the users" impression will *e of your department as a cohesie unit where groups communicate among one another! o
(trie
for
9CC
percent
uptime
=ontinuous uptime may seem like an o*ious goal for anyone who runs a computer system, *ut keeping it in your mind at all times is a real challenge! 0ow, eeryone who"s worked with 0) for any length of time knows that 0) serers need re*oots eery so often, *ut a really good 0) administrator should try to minimi#e this need! eep your serers simple, assign them only the tasks you need them to perform, and don"t clog up the system with additional serices or applications unless they"re a*solutely essential! )here"s a saying that a person"s *ody is his or her templeK make temples out of your production *o4es! ?ae fun with your test serers, *ut when you hae a spare moment, spend it working to make your 1indows 0) serers as fault-tolerant as they can *e! /s I mentioned earlier, many administrators don"t hae a pro*lem spending hundreds or thousands of dollars to make their hardware *ulletproof! +ut in the hours necessary to make your operating systems and software just as *ulletproof!