How to reboot a server. Need advice.
Jeff Shultz
sucks and isn't BOFH'ish, but taking down a server will require some
time sacrifice on your part.
On Sun, 13 Dec 1998 04:15:13 GMT, Jeff Liebermann wrote:
:>I have a protocol problem. Every time I shutdown a server for
:>maintenance, someone always complains. What is the correct proceedure
:>for informing the GUM (great unwashed masses) that at precisely 5:00PM,
:>the sacred server will be down for about 10 minutes while I untangle the
:>cables from the furniture?
:>
:>I've tried everything from tactful diplomacy to brutality in announcing
:>that the server will be down. It never fails; at the appointed hour,
:>someone will always be "just finishing up something important" and need
:>"just a few more minutes". Naturally, I'm not informed when they're
:>done. They also don't bother to logout and prefer to just walk away. I
:>went to extremes at one customers and printed announcements which I
:>stuck on all the monitors in sight. It didn't work. There were a half
:>dozen clueless morons who simply walked away, leaving applications,
:>databases, email, and/or files open. If I just dump them, they're sure
:>to lose some data. By the time I get them all safetly logged off, it's
:>30 minutes later and other users are screaming.
:>
:>Even if I manage to get the server down without violence, someone will
:>always appear within 60 seconds and ask "Whenz-da-servah-gonna-be-up"?
:>It's like their world just stops when the server goes down.
:>
:>So, what works? How do I shutdown a busy server without violence, data
:>loss, or a nervous breakdown?
:>
:>
:>--
:>Jeff Liebermann 150 Felker St #D Santa Cruz CA 95060
:>(831)699-0483 pgr (831)426-1240 fax (831)336-2558 home
:>http://www.cruzio.com/~jeffl WB6SSY
:>je...@comix.santa-cruz.ca.us je...@cruzio.com
Jeff Shultz
If Life is a freeway, why can't I find an onramp?
Jeff Liebermann
Sentinel
----------
>From: je...@comix.santa-cruz.ca.us (Jeff Liebermann)
>Newsgroups: alt.tech-support.recovery
>Subject: How to reboot a server. Need advice.
>Date: Sat, Dec 12, 1998, 11:15 PM
>
[snip]
>Even if I manage to get the server down without violence, someone will
>always appear within 60 seconds and ask "Whenz-da-servah-gonna-be-up"?
>It's like their world just stops when the server goes down.
>
>So, what works? How do I shutdown a busy server without violence, data
>loss, or a nervous breakdown?
>
Get some backing from above. Tell whoever has authority over everyone else
to back up a statement that you make by putting his name on it. I would
also use the terms (am not responsible for lost data) and no reprieves
unless the Governor (higher authority) authorizes it, no exceptions.
:)
Paul Raj Khangure
: I have a protocol problem. Every time I shutdown a server for
: maintenance, someone always complains. What is the correct proceedure
: for informing the GUM (great unwashed masses) that at precisely 5:00PM,
: the sacred server will be down for about 10 minutes while I untangle the
: cables from the furniture?
0) Notify everyone of time server will be down.
1) Divert your phone to the weather report number.
2) Lock your door / machine room door.
3) Take down server.
4) Bring server back up.
5) Have a few drinks.
6) Unlock your door / machine room door.
7) Go home.
Paul Raj Khangure
--
I stayed up all last night playing poker with tarot cards.
I got a full house and four people died.
I don't suffer from insanity, I enjoy every minute of it.
SteveD
>I have a protocol problem. Every time I shutdown a server for
>maintenance, someone always complains. What is the correct proceedure
>for informing the GUM (great unwashed masses) that at precisely 5:00PM,
>the sacred server will be down for about 10 minutes while I untangle the
>cables from the furniture?
[Tales of GUM snipped]
I find that one of the better approaches is to have a large sign
(cardboard if necessary, wood if possible) that says
"As you have already been informed, the server will be down from 5.00 to
5.10pm. It is your responsibility to arrange for the safety of your work
during this time. If you do not know how to do this, please contact the
Training area on extension XXXXX."
Arrange for your manager or whoever in the food chain is willing to take
responsibility (as high up as you can go) to sign a document that
determines exactly what warning you should be giving in the event of
necessary maintenance. To avoid them making things up, present them with
a tick-box document, with things like "Email everyone on the morning of
the day" "Remind everyone about it ten minutes before the scheduled
time" etc. Make sure that none of the options listed will cause you too
much hassle. Get them to sign and date the document. Give them a copy.
Give your manager a copy (assuming they're not the same person). Keep
the original blank one for when the policy-signer decides to change the
policy (they will).
After jumping through the appropriate policy hoops on the day, take your
phone off the hook at 4.50pm or whenever the last warning is given. Walk
to the server and place the sign in a prominant position where you can
point to it without speaking. Down the server at 5 and perform necessary
tasks. If lusers appear, point to the sign.
If anyone complains, refer them to your manager. Your manager can then
refer them up the food chain to the policy-maker. Assuming they get that
far, you always have the blank document prepared for any changes they
may want to make, and the cycle starts again. You might also want to
have an official-looking document that says "Authority to set policy on
downtime for server <SERVERNAME> delegated to ________ on <date>" for
when the policy-maker gets tired of lusers and delegates it down the
line...eventually this authority may delegate to you, and you will then
be officially authorised to tell the lusers to go jump.
Appropriate preparation means you can always pass the buck.
If the policy-maker asks why the server needs so much downtime, say it's
because it's a model <whatever> and the new, improved model would cost
<whatever it would cost for such a beast plus extras plus 10%>. Act
enthusiastic about the prospect of getting this new model.
Hope this helps.
--
Steve in .au
Pre-guiding policy in Politikopolis
Top Jimmy
Liebermann) wrote:
>I have a protocol problem. Every time I shutdown a server for
>maintenance, someone always complains. What is the correct proceedure
>for informing the GUM (great unwashed masses) that at precisely 5:00PM,
>the sacred server will be down for about 10 minutes while I untangle the
>cables from the furniture?
>
Well, when I was an AS400 Sysop, it was easy. "SNDBRKMSG->Alt+F4->The
system will be taken down for maintenance in 10 minutes. Please log
off. Thank you.-> enter". After 10 mintues, you lock out all the
regions on the WAN, and then go for it. Of course, that didn't solve
the problem with the lusers who either couldn't read or didn't
understand the system availability schedule. Without fail, they would
call 2 hours into a weekly NONSYS that took 3 hours...
:RING
Me: IS, BAS400OFH speaking.
Luser: When will the system be back up? Is there something wrong?"
Me: <thinking>Yeah, there's something wrong, I work the graveyard
shift, so I get stuck taking all the help desk calls from back east
because the help desk isn't here yet... oh what a cruel twist of
fate![1] Shoot me now!</thinking> No, the weekly backup always takes
this long, the system is not scheduled to be available for another
hour. Didn't you get the system availability schedule?
Luser: No.
Me: It was sent out via Notes *twice*.
Luser: Oh.[2]
GOTO RING
[1] I left a tech support job for operations so that I didn't have to
deal with lusers like this. It didn't work, though, I was doing so
much Bobbing, it didn't make a difference. Now I am a PC Technician
for a network solutions/PC repair/web hosting/custom hardware company.
Bring on the lusers!
[2] Lusers are too busy mailing each other chain letters and
tasteless jokes to pay attention to important e-mail.
------------------------------------------------------------------
"Now when I was a young boy, at the age of 5, my mother said imma
be, the greatest man alive." - McKinley Morganfield
Top Jimmy - http://www.inficad.com/~cdboy
------------------------------------------------------------------
Jacob Haller
> So, what works? How do I shutdown a busy server without violence, data
> loss, or a nervous breakdown?
My advice would be to go to the actual machine room (where hopefully
there isn't a phone, or if there is a phone few people know the number),
do what you need to, and go home.
Data loss by people who are not following outlined procedures is not
your problem. Hopefully.
-jwgh
Bram Smits
>I have a protocol problem. Every time I shutdown a server for
>maintenance, someone always complains. What is the correct proceedure
>for informing the GUM (great unwashed masses) that at precisely 5:00PM,
>the sacred server will be down for about 10 minutes while I untangle the
>cables from the furniture?
Tacnuking their site usually seems to get their attention. Anything less
is just a waste of breath.
>I've tried everything from tactful diplomacy to brutality in announcing
>that the server will be down. It never fails; at the appointed hour,
>someone will always be "just finishing up something important" and need
>"just a few more minutes".
That's why you disable logins and do a few wall's (or whatever the
equivalent is on your platform of choice) that the system is going down
starting at about T-15minutes. At T+1seconds, you start kicking them off
the system (or just down the system if you don't have apps running that
need the lusers kicked off).
>dozen clueless morons who simply walked away, leaving applications,
>databases, email, and/or files open. If I just dump them, they're sure
>to lose some data.
Their problem, not yours.
>By the time I get them all safetly logged off, it's
>30 minutes later and other users are screaming.
The other option is always the mutua-LART: Get on the PA system and
announce that "for maintenance reasons, system X[1] will be unavailable.
Because the following lusers <list names> failed to log off as requested
in the [memo|poster|motd|mail|messages], this will take <insert large number>
[minutes|hours] more than neccesary. If you are inconvenienced by this, I
suggest you take it up with them. The lusers concerned are at the following
locations: <list luser and most likely location>."
Then sit back and watch the fun on the security cams.
>Even if I manage to get the server down without violence, someone will
>always appear within 60 seconds and ask "Whenz-da-servah-gonna-be-up"?
>It's like their world just stops when the server goes down.
Turn off your cellphone and beeper, and unplug your desk phone. Barricade the
doors to systems central, and lower the shades on all windows facing hallways.
If you're feeling generous, paste a note to the door of systems central
informing them that the server will be up in 10 minutes once all the
idiots have properly logged out. If you're feeling evil, put a monitor
behind the door listing the idiots still logged in and where they are.
>So, what works? How do I shutdown a busy server without violence, data
>loss, or a nervous breakdown?
Without violence: No fscking way. You just have to work on directing that
violence. I don't know what business hours your company keeps, but you might,
in a moment of weakness, consider doing shutdowns outside business hours, or
at least at the end of business (1730 or 1800 hours).
Without data loss: Just make sure you terminate all sessions to databases
and the like that are sensitive to such things. What do you care if some
luser data gets lost ?
Without a nervous breakdown: Just don't give a toss about the previous two
points.
v__
<"___\____ Bram 'mouser' Smits.
[1] for whatever value of "system X" your lusers understand.
--
I think, therefore I think I am
Tim Franklin
wrote:
>I have a protocol problem. Every time I shutdown a server for
>maintenance, someone always complains. What is the correct proceedure
>for informing the GUM (great unwashed masses) that at precisely 5:00PM,
>the sacred server will be down for about 10 minutes while I untangle the
>cables from the furniture?
>
It really depends on how far in advance you know. A week or more is generally
good, but if you make it too long, they'll forget when. A week is probably
about right, if you can engineer it.
[snip terror stories]
>
>So, what works? How do I shutdown a busy server without violence, data
>loss, or a nervous breakdown?
>
Mail[0] to everyone affected. CC to PHBs as far up the chain as far as you can
get away with, preferably as far as the original founder of the company[1].
Write clear simple instructions, in short sentences as to just what you mean by
"log off and don't leave any apps running". Use small words only. Repeat. A
lot. Instructions a senile baboon could follow is what we're going for here.
At the top and bottom, write in capitals, and again in short simple sentences
that if they do not follow the instructions in this mail, to the letter, before
the appointed time, they *will* lose data, and it will be *their* fault, not
yours.
Repeat mailing as the shutdown time approaches with things like 'REMINDER -
URGENT SHUTDOWN - POTENTIAL DATA LOSS' at the top.
It does the job *fairly* well for me, although we did have the moron who didn't
understand workstation shutdown. Most bank holidays, the sparkies come in and
"do stuff" with the mains wiring. This involves them repeatedly switching on
and off the power to various areas of the building, which is not good for any
machines with a switch still in the 'on' position. To this effect, we run a
script[2] that shuts down every workstation, then we go around the building and
physically power down each of about 600 machines.
Anyway, this one uber-luser was insistant that he had to keep working. We told
him he couldn't[3] and shut him down. A couple of cubes down, we hear clicking
- he's powered back up, and he's typing again. Shutdown. More cubes.
Clicking. We simply took his power cables away in the end. Power *that* back
on, you fornicating chimpanzee! *sigh*
I'll go now...
Regards,
Tim.
[0] Or whatever internal messaging system you have. If you have several, use
the one that has external non-business related access - they read that one.
[1] I have a lovely spoof SOP on my desk that requires the approval of the late
[snipped] for every workstation I install. The scarey thing is that's probably
the only thing in it that isn't accurate to the way the installs *really*
happen.
[2] Many hundred lines of perl. Don't ask, it's *not* mine.
[3] This is about an hour after the designated shut-down time, we've already
been round three other floors.
--
Tim Franklin | "Progress is the process by which Usenet
| has evolved from smart people in front of
Home : t...@leaphome.pelican.org | dumb terminals to dumb people in front of
FidoNet : Tim Franklin@2:257/15 | smart terminals." -- Obscurity, 1997
Chris King
/removethis/pot.net> writes
>Correct procedure:
>
>1) Post clearly and publicly that the server is going down at 5:00. Make
>sure that the news's been thoroughly spread.
When I post warnings of downtime, I also mention *which* services will
be unavailable, and an estimate of how long I expect the system to be
down. Don't try to be exact with this, allow yourself time to correct
minor faults or roll back if you have to.
>2) At 5:00, shut down the server.
On the dot - if someone is still logged in despite being warned, that's
their problem.
>3) At 5:05, explain to the whiners that you'd warned them, and that when
>you say 5:00, you mean 5:00.
You have to be VERY strict on this one. If you delay shutdown to satisfy
one person, you'll have other people asking you to do the same and next
thing you know, it's the wrong side of last orders :-(
>4) Repeat steps 1 and 2 as frequently as possible until people get the
>hint.
One problem you may have is that if you post the downtime info too long
before the event, people keep phoning up saying "I forgot !" when you
point exactly why the system is down.
Chris
--
Chris King
ch...@csking.demon.co.uk
http://www.csking.demon.co.uk
Quentin Stephens
>I have a protocol problem. Every time I shutdown a server for
>maintenance, someone always complains. What is the correct proceedure
>for informing the GUM (great unwashed masses) that at precisely 5:00PM,
>the sacred server will be down for about 10 minutes while I untangle the
>cables from the furniture?
Do a walk-round of the department that morning, email everyone. At 16:50 send
a broadcast message and prevent new logins. At 16:55 check who's logged in
and visit or phone them. At 17:05 down the server. And always say that the
server will be unavailable for half an hour. You're setting an expectation
and then looking great by exceeding it.
qts
Usenet readers please reverse the elements of my given address to get my real one
void
wrote:
>
>I stayed up all last night playing poker with tarot cards.
>I got a full house and four people died.
>
>I don't suffer from insanity, I enjoy every minute of it.
Do you have something against attribution? For at least one of those
quotes, the author's identity is trivially discoverable.
--
Ben
"You have your mind on computers, it seems."
hiroshi
says...
> On 13 Dec 1998 08:45:42 GMT, Paul Raj Khangure <p...@opera.iinet.net.au>
> wrote:
> >
> >I stayed up all last night playing poker with tarot cards.
> >I got a full house and four people died.
> >
> >I don't suffer from insanity, I enjoy every minute of it.
>
> Do you have something against attribution? For at least one of those
> quotes, the author's identity is trivially discoverable.
>
>
seem to recall the second one being Groucho Marx.
Hiro
hiroshi
teknique@honey/removethis/pot.net says...
> Jeff Liebermann wrote:
>
> >I have a protocol problem. Every time I shutdown a server for
> >maintenance, someone always complains. What is the correct proceedure
> >for informing the GUM (great unwashed masses) that at precisely 5:00PM,
> >the sacred server will be down for about 10 minutes while I untangle
>
> Correct procedure:
>
> 1) Post clearly and publicly that the server is going down at 5:00. Make
> sure that the news's been thoroughly spread.
> you say 5:00, you mean 5:00.
> hint.
> Tek the Glazed
>
The Company I slave for uses voice mail very heavily, we had a major
outtage scheduled the other night so it was arranged for one of the high
level managers to send a voice mail to the entire company informing them
of this, this was sent twice in the 2 days preceeding the outtage,
naturaly I'm stuck on graveyard shift when this occurs.
12:00am, Server is taken offline and the the admins get to work
12:03am, my phone rings,
(luser) I can't log into X
(Me) Thats because server y is offline an upgrade, didn't you get the
voicemail
(l) yes..., but how would that affect X, it runs on server z
(m) but it needs the database, which runs off server y, the voicemail
specificly said that X would be unavailable.
(l) but I need to use it urgently, when will it be back up
(m) 6am [1]
(l) What!!!!!, why do you always schedule these things for the most
inconvenent times, what am I supposed to do now? [2]
(m) It's 12:05 on a sunday morning, would you like us to do it at midday
on a tuesday?
This went on for another 5 minutes, I then had a another call allmost
exactly the same..
Down, not Across.....
[1] They actualy hoped to have it back up at 3 but I wasn't telling him
that
[2] Get a life maybe, not that I know what one of those is [3]
[3] I start work at 8am on Jan 1st, and one of my friends has just
announced what will be the mother of all new year's eve parties, I even
get to work right through Xmas as well
Hiro
mikeh...@hotmail.com
Liebermann) wrote:
>I have a protocol problem. Every time I shutdown a server for
>maintenance, someone always complains. What is the correct proceedure
>for informing the GUM (great unwashed masses) that at precisely 5:00PM,
>the sacred server will be down for about 10 minutes while I untangle the
>cables from the furniture?
>
>I've tried everything from tactful diplomacy to brutality in announcing
>that the server will be down. It never fails; at the appointed hour,
>someone will always be "just finishing up something important" and need
>"just a few more minutes". Naturally, I'm not informed when they're
>done. They also don't bother to logout and prefer to just walk away. I
>went to extremes at one customers and printed announcements which I
>stuck on all the monitors in sight. It didn't work. There were a half
>dozen clueless morons who simply walked away, leaving applications,
>databases, email, and/or files open. If I just dump them, they're sure
>to lose some data. By the time I get them all safetly logged off, it's
>30 minutes later and other users are screaming.
>
>Even if I manage to get the server down without violence, someone will
>always appear within 60 seconds and ask "Whenz-da-servah-gonna-be-up"?
>It's like their world just stops when the server goes down.
>
>So, what works? How do I shutdown a busy server without violence, data
>loss, or a nervous breakdown?
Announce 15 minutes or so before hand, "The server will be shut down
for maintainence purposes in X minutes at X a.m/p.m. NO IFS ANDS OR
BUTS. I would politely suggest that anyone wishing to save ANYTHING
they are currently working on, DO IT NOW. The server will be down for
APPROXIMATELY ONE (whatever unit of time). Thank You. PLEASE SAVE
ANY WORK YOU WISH TO SAVE NOW."
As for important, remember Muttleys' first rule of tech support:
Never believe the sons of bitches are telling you the truth to begin
with.
Muttleys' second rule of tech support: Even if you do it exactly on
time and exactly as announced, somebody will always be pissed. ITS THE
LUSER WAY!
11 more drinking days 'til Xmas,
Muttley
Paul Raj Khangure
Yes
Steve
university's main public Unix box, which also serves filespace for two other
key servers. Between the three, there are never less than 150 logins at any
time. According to our semi-BOFH, proper system maintainance begins with
architecture:
1. Put the server room in a controlled part of the building. Choose a place
that isn't easily guessed, such as in a part of the basement accessible only
through a semi-hidden stair. Don't list its location in the building
directory.
2. Have only one phone in the server room. This phone number shouldn't be
publicised or "belong" to anyone. It'll be off the hook during the procedure,
of course.
3. Starting a week prior, include the date and time of the maintainance in the
login message.
4. Schedule a time when no official support outfits are open. Easy here, where
the helpdesk is a 9-6 operation. The operators here prefer to do it in the
morning, so as to finish right before the business day starts.
5. Take your office phone off the hook, and disable voicemail. Go to the
server room.
6. Do your business. Don't worry about lusers losing data. It's their own
fault.
7. Ignore all complaints. Be an asshole about it. They saw the message.
Elrond Hubbard
Jeff Liebermann <je...@comix.santa-cruz.ca.us> wrote:
>I have a protocol problem. Every time I shutdown a server for
>maintenance, someone always complains. What is the correct proceedure
>for informing the GUM (great unwashed masses) that at precisely 5:00PM,
>the sacred server will be down for about 10 minutes while I untangle the
>cables from the furniture?
<snip>
>So, what works? How do I shutdown a busy server without violence, data
>loss, or a nervous breakdown?
Be ruthless. Give them all fair warning, then just do it. If they lose data,
tough. Once it happens a few times they'll get the idea that you mean
business.
--
Elrond Hubbard, lbe...@mnsi.net
I may speak for my employer, but I post for myself.
Jeff Liebermann
>I have a protocol problem. Every time I shutdown a server for
>maintenance, someone always complains. What is the correct proceedure
>for informing the GUM (great unwashed masses) that at precisely 5:00PM,
>the sacred server will be down for about 10 minutes while I untangle the
>cables from the furniture?
OK, I'll try to summarize the suggestions and why they won't work.
1. Sufficient advanced warning:
One of my customers is a hospital. Taking down the ER computer requires
careful planning. They generally condescend to give me either that
2-3am or the 5-6pm slot for maintenance. Two weeks notice is usually
required. No matter how hard I try, there's always some major emergency
in the middle of system maintenance.
Incidentally, the reason 5-6pm is an ER slack time is that all the
doctors offices are just closing up and the ambulances have yet to drag
in the latest commute casualties. The hospital office staff just left.
The parents have just arrived home to notice that baby is running a
fever. Moving the downtime to 6pm would be a disaster in a hospital.
Email is an inadequate method of distributing information because many
of the users do not use or have email. While corporate America now runs
on email, much of the service sector sees no need for everyone to have
email.
2. Develop an "I don't care attitude":
I did this and it doesn't work. The average user cares even less,
bordering on retaliation. A typical user that always logs off at
exactly 5pm will magically "forget" to logout only on the day I need to
have the server down. Any change in their routine (such as scheduled
downtime) results in instant clueless behavior. My big fear is that if
I act callous and indifferent, they will escalate their level of
accidental ineptitude and start a war which I am certain to lose.
There's more of them than me.
Making myself unavailable doesn't work because I charge for getting
pages and calls. Every time the pager goes off with a customers phone
number, I think "$50 minimum NET 10". I love my pager. I even have a
spare. Charging for pages and calling back right away is my main source
of business. I tell everyone "call me BEFORE you break it trying to fix
it." I don't even have a voice phone in the office (because I'm never
there during business hours anyway).
3. Get management involved:
I can't get support from my boss because I'm my own boss. In my line of
contract work, remaining invisible is a requirement for keeping the
contract. I'm there because the boss could not keep his own computers
alive and could not keep his users in line. I'm there so the boss can
concentrate on more important things, whatever they may be. In short,
moving the problem up a notch is suicidal. It's my problem and will
remain my problem.
The one time that I did attempt to get downtime scheduled in an
organized manner, the high command decided that *ALL* undesireable and
inconvenient tasks were to be scheduled for a specific day. When I
arrived, I was greeted by the carpet cleaners whos noxious chemicals
nearly gassed me, and a mob of wiring guys who also setup shop in the
server closet. There were also other service companies trying to do
upgrades at the same time. When the painters arrive, he took one whiff
and just left. This is what happens when you get management involved in
scheduling.
4. Elevate the urgency and method of notification:
Announcing that the world will end if they don't logout now, is subject
to the "Chicken Little Effect" where users become callous in the event
of real emergencies. Exessive repetition of the same message will only
cause the user to ossify in a state of frozen confusion. I save such
messages as "The World Ends at 5PM if you don't logout now" for special
occassions like my accidently killing init, which does an impressive job
of trashing any Unix box. Messages to the screen using "wall" tend to
be a waste of time because the typical user will think their MS Word
document went insane, and then call support asking for an interpretation
of the "error" message. I've tried it both ways and found that fewer
users will logout after being notified on-screen, than without.
Using a public address system doesn't work in buildings that don't have
them, hospitals, and companies with a large number of remote offices.
There are also companies that guard the PA system telco code more
effectively than the administrator/root passwords. Only the chosen few
may use the PA system. I can't wait for VoIP where shutting down the
network also takes down the phone system.
It seems that there's no standard way of shutting down a server for
maintenance. Most of the suggestions are system or company specific.
It would be easy if the operating system and application scribblers
would provide for a signal that magically saves the work, exits the
application, and logs the user out. I've yet to see this, so psychology
and brutality appear to be the only options.
Scott Brown
wrote:
>In article <36763a79...@news.ricochet.net>,
>Jeff Liebermann <je...@comix.santa-cruz.ca.us> wrote:
><snip>
>>So, what works? How do I shutdown a busy server without violence, data
>>loss, or a nervous breakdown?
>
>Be ruthless. Give them all fair warning, then just do it. If they lose data,
>tough. Once it happens a few times they'll get the idea that you mean
>business.
Yeah, *right*.
It's not quite the same thing, but for a while I had the unenviable
job of closing a college computer lab at night, which mainly involves
getting everyone booted off the machines and out the door. I'd make
the announcement at 20 and 10 minutes to close, and when closing time
arrived there were inevitably two or three people still pounding away
at the keyboard.
Early on I tried to be polite, asking them to please leave so I could
go home. After a week or so of this approach not working, I adopted
the technique of simply yanking the power cable out of the wall
socket[1]. Although it had the desired effect of getting people to
leave, the lusers never did learn that when I said "it's time to go",
I meant it. Face it: if they were capable of absorbing clues, they
wouldn't be lusers, would they?
[1] Yes, I know, but (a) they're win95 boxes, and (b) I wasn't
responsible for fixing them.
Steve
> It's not quite the same thing, but for a while I had the unenviable
> job of closing a college computer lab at night, which mainly involves
> getting everyone booted off the machines and out the door. I'd make
> the announcement at 20 and 10 minutes to close, and when closing time
> arrived there were inevitably two or three people still pounding away
> at the keyboard.
I have a similar task on Sunday nights, but policy around here makes it
easy. Mind you, this is what I'm told to do. I make the usual announcements up
until closing. Two minutes after closing I make a slightly firmer request.
Five minutes after closing I announce that the police will be called to remove
anyone who does not leave immediately. I've never had to carry out the threat.
Chris Adams
>the technique of simply yanking the power cable out of the wall
>socket[1]. Although it had the desired effect of getting people to
[...]
>[1] Yes, I know, but (a) they're win95 boxes, and (b) I wasn't
>responsible for fixing them.
There's good use for WinNuke and friends. "Gee, it's a shame you didn't log
out by the time you were told to. Your savefile might not have gotten
corrupted..." Any resemblance between the corrupted savefile and the results
of an errant "cat /dev/random > /lusers/luser/termpaper.doc" is purely
coincidental.
--
Why must every new release confirm my belief that Bloody Stupid Johnson
is now employed by Microsoft?
Bob the Sane
that bleedin' server and let the poor sumbitches rot in fscking hell.
Your server is too good for them, and they only use it by the grace of
God (root, that is).
Bohica
>a broadcast message and prevent new logins. At 16:55 check who's logged in
>and visit or phone them. At 17:05 down the server. And always say that the
>server will be unavailable for half an hour. You're setting an expectation
>and then looking great by exceeding it.
I agree. You can never go wrong by using the Scotty method of time estimation.
Allow enough time to get a quick game of Quake (or whatever) er, I mean, intense
network testing, in....
Jeff Liebermann
<s...@mardlin.oc.ku> wrote:
>Do a walk-round of the department that morning, email everyone.
Morning? Whazzat? I get started at about noon and work till midnight.
Computers are best done under cover of darkness.
A walk-around would be a bit too much exercise. Some of the users are
scattered around acres of "campus", in multiple buildings, or in remote
offices at the end of lease lines. I don't even know where most of them
hide. Some companies reorganize so often that I have trouble finding
the building much less the users. One machine averages about 200 users
which would take all day to find and inform.
To make things even more interesting, I'm often not on-site for minor
software upgrades and fast fixes. I do these via a dialup connection
(using ssh or RAS) making a service call un-necessary. That eliminates
the walk around.
Email to everyone is also a problem. Ignoring the users that don't have
email (there are plenty of those), the "everyone" alias tends to be
innundated with management announcement, internal spam and changes in
punctuation policy statements, which leads to users religiously ignoring
anything from the "everyone" alias. One astute user wrote a good
description of how to use Eudora filters to trap email addressed to
"everyone" and promptly emailed it to the "everyone" alias.
Ryan Tucker
<je...@comix.santa-cruz.ca.us> spewed:
>would provide for a signal that magically saves the work, exits the
>application, and logs the user out. I've yet to see this, so psychology
SIGTERM is /supposed/ to result in a clean shutdown...
... and I was supposed to just stay up all night to take someone to the
airport this morning. -rt
--
Ryan Tucker <rtuck...@ttgcitn.com> http://www.ttgcitn.com/~rtucker/
GSM/VM/Fax: +15157712865 Box 57083, Pleasant Hill IA 50317-0002
Jeff Liebermann
rtucker+f...@katan.ttgcitn.com (Ryan Tucker) wrote:
>On Mon, 14 Dec 1998 18:52:26 GMT, Jeff Liebermann
><je...@comix.santa-cruz.ca.us> spewed:
>>would provide for a signal that magically saves the work, exits the
>>application, and logs the user out. I've yet to see this, so psychology
>
>SIGTERM is /supposed/ to result in a clean shutdown...
Want a list of applications and situations where this doesn't
work? The application may get cleanly shut down, but the users
work certainly doesn't get saved. In databases, I don't want the
"work" saved unless the user has at least looked at the screen to
be sure it correct. In cases of an emergency shutdown, where the
system has gone nuts, the last thing I need is to have every
application scribbling all over its data files as the system goes
down. Most programmers have opted for the safe approach and
decided to not write anything on the way down. Hmmm... at a
second thought, perhaps it's easier to remind the users to save
their work than it is to clean up the mess after a shutdown.
>... and I was supposed to just stay up all night to take someone to the
>airport this morning. -rt
You got it easy. My truck just blew a head gasket or cracked the
head (again). Time to get greasy.
Ryan Tucker
<je...@comix.santa-cruz.ca.us> spewed:
>>SIGTERM is /supposed/ to result in a clean shutdown...
>
>Want a list of applications and situations where this doesn't
>work? The application may get cleanly shut down, but the users
I can probably generate a list of my own. Hence "supposed". The
definition of a clean shutdown does vary, as well.
>>... and I was supposed to just stay up all night to take someone to the
>>airport this morning. -rt
>
>You got it easy. My truck just blew a head gasket or cracked the
>head (again). Time to get greasy.
I decided it was time to rotate the tires on the Neon, given that my front
ones are nearing retirement and my back ones are good enough to pass as
new.
Well, I do have aftermarket wheels, which were installed by a local tire
shop many months ago. They have a nice little plate covering the nuts,
held on with a screw.
The right front was fine... took a bit of powertool, but tire management
is an industry where using loud impact wrenches is supposed to make the
customer feel safe and secure, despite their rotors very likely becoming
warped to hell and back. No incredibly big deal.
The left front was a bitch. The screw was very well rounded by the time
the power tools were exhausted. Fsck. Get out the little thingy which is
designed to grip in situations where you're in over your head, work on it
for a few minutes, and the screw comes out. It's stripped like a tabletop
after a cheap vodka spill, but it's out.
The left rear goes much the same.
The right rear, OTOH, scoffs at any such attempt. Hammer away, and it
still stays there. Stripped like... well, stripped like nothing that has
ever been stripped before.
Drench it in WD-40, park the car in the garage, and ground it for the
night... we'll see what happens in the morning. If worse comes to worse,
I may have to impact-wrench the tire guy's testicles. -rt
Tim Franklin
wrote:
[snippage]
>It would be easy if the operating system and application scribblers
>would provide for a signal that magically saves the work, exits the
>application, and logs the user out. I've yet to see this, so psychology
>and brutality appear to be the only options.
>
Isn't SIGTERM supposed to cause things to exit in as polite a fashion as
possible? If not, just what *is* its function when we have SIGKILL?
Exceptions, of course, for clueless app vendors and boxen that don't run the One
True OS[0], make this next to useless :(
Regards,
Tim.
[0] Alright, the Many True OS Flavours. I'm not going to get into a "your *nix
sucks harder than my *nix" holy war.
Jeff Liebermann
Tucker) wrote:
Warning: Topic drift.
>I can probably generate a list of my own. Hence "supposed". The
>definition of a clean shutdown does vary, as well.
The definition of a clean shutdown is easy.
1. The phone doesn't ring.
2. My pager doesn't go off.
3. The server reboots without any error messages or forced fsck.
4. Nothing is lost, trashed, totaled, vaporized, or added.
5. The NFS automounter, BDC, DFS and other dependent server services on
other servers don't loose touch with reality.
6. No screams are heard from the office area.
7. Nobody even notices.
I've come close, but never all the above.
>I decided it was time to rotate the tires on the Neon, given that my front
>ones are nearing retirement and my back ones are good enough to pass as
>new.
I don't believe it.
Nobody rotates their tires.
Nobody does backups.
Nobody pays for shareware.
Nobody trusts computer documentation.
>night... we'll see what happens in the morning. If worse comes to worse,
>I may have to impact-wrench the tire guy's testicles. -rt
Remind me not to let you anywhere near my trucks. I use some kind of
anti-sieze goo on everything exept alloy wheels. Incidentally, the guy
with the impact wrench has instructions that it's better to torque the
lug nuts on too tight than to risk litigation when the wheel falls off.
When I was playing RF (radio) engineer, my standard question to
prospective technicians was "do you work on your own car"? If yes, then
I would ask for a few details. Generally, anyone that does their own
maintenance makes a better technician as they have already mastered the
fine arts of duct-tape application, big hammer adjustment techniques,
baling wire support construction, creative part substitution, and
applied profanity. The overhead is also less as they've mastered the
conversion of any convenient tool into a hammer.
I've often wondered why the cube-farm or desk pilot variety of
tech-support is so frustrating. It's been many years since I did this
kind of work. My guess is that it's like giving flying lessons without
ever being allowed to fly. It must be terribly frustrating.
I've futilely suggested that programmers be forced to spend one day per
month answering phone support questions about their wonderful product as
penitence for their sins. Similarly, methinks that inside tech support
people should go out in the field and deal with the realities of field
service. Then, they can experience things like:
"What? How do you shut down the server? I can't hear you over
all the beeping! What?? Louder please".
Like Windoze 95 installs that demand a blank floppy. Have you ever
tried to find a blank floppy in a big busy office? Good luck. Every
floppy in sight has something unidentifiable on it usually without
decypherable lables. It take hours to find one that can be erased. Or
how about RTFM? Never mind reading the manual, just try to find the
manual.
"Joe borrowed it 3 months ago. Haven't seen it since".
Same with original floppies and reinstalls:
"What do you mean you can't find Disk #41"?
Server support is even more entertaining. Step number one is to find an
empty AC power outlet. This can take anywhere between 30 seconds and
all day.
"Does 5 power strips in series violate some kind of NEC code
or ordinance"?
Maybe some field experience or auto repair might help make tech-support
a better hobby.
Anyway, I lucked out. Just a blown head gasket. No damage to the head
or block. An easy repair if I can find a head gasket for a 16 year old
diesel engine. The only problem is that I didn't wash my hands too well
before I starting typing this rant. Now, I gotta degrease the keyboard.
Yech.
Jeff Liebermann
Franklin) wrote:
>Isn't SIGTERM supposed to cause things to exit in as polite a fashion as
>possible? If not, just what *is* its function when we have SIGKILL?
Oh swell. A tech support question in a tech-support newsgroup. What is
this newsgroup coming to? Support questions are everywhere. I made the
mistake of wearing a Cruzio (a local ISP) T-shirt to a street fair. I
got stopped a dozen times with "You must work for Cruzio and I have a
computer problem". I should be like the smarter users and play dumb.
It attracts fewer support questions.
See:
http://www.sco.com/cgi-bin/ssl_getmanpage?signal+S+OS5
for a handy list of this weeks signal(S) standard. Note the variations
in support under XPG3, POSIX and ANSI. SIGTERM(15) closes the
application if the app feels like closing. If it's waiting for input,
it will generally hang around until it gets fed before shutting down.
SIGTERM is also trappable which makes some applications unkillable.
The heavy hammer is SIGKILL(9) which cannot be trapped, but also is
particularly graceful about the way it terminates an application.
SIGTERM is quite graceful, while SIGKILL is brutal.
In both cases, we're talking about shutting down the application. No
provisions are made for saving the data or work. When the application
shuts down, the work is usually lost. Some programs that use
transaction files are reasonably smart about recovering from a reboot.
However, the bulk of the applications I've seen just blow away the work.
Some are stupid as they use a work file but overscribble the workfile
with a blank file with the same name on restart.
What I was looking for was a way to save literally the entire user
environment. That would be the shell, X11 environment, env variables,
application memory image, data file image, all tmp files, etc and reload
it on reboot. Something like the "suspend" feature found on many
laptops. The main advantage is that I could reboot a server without
much consideration for what the clueless users are doing. From their
point of view, the "system" froze for a few minutes while it did a
reboot. When the system comes back up, all the logins are restored, and
life continues uninterrupted. I vaguely recall the Tandem's Non-Stop
Kernel actually did this.
I haven't really worked out all the implications of this pipe dream. I
suspect that there will some complications and difficulties. However,
if it could be done, my server reboot notification problems would be
mostly solved.
Ryan Tucker
<je...@comix.santa-cruz.ca.us> spewed:
>>I decided it was time to rotate the tires on the Neon, given that my front
>>ones are nearing retirement and my back ones are good enough to pass as
>>new.
>
>I don't believe it.
> Nobody rotates their tires.
Time slips by quickly. I've been basically abusing my car for the last
year, and now it's time to make up for it. I need new tires anyway; the
current ones might as well be bald on wet pavement.
>>night... we'll see what happens in the morning. If worse comes to worse,
>>I may have to impact-wrench the tire guy's testicles. -rt
>
>Remind me not to let you anywhere near my trucks. I use some kind of
>anti-sieze goo on everything exept alloy wheels. Incidentally, the guy
>with the impact wrench has instructions that it's better to torque the
>lug nuts on too tight than to risk litigation when the wheel falls off.
This is the first time I've personally tried to remove the wheels...
needless to say, given that I've yet to even get to the lug nuts on one
wheel because of overtightening, I'd much rather torque things the Right
Way [tm].
>When I was playing RF (radio) engineer, my standard question to
>prospective technicians was "do you work on your own car"? If yes, then
>I would ask for a few details. Generally, anyone that does their own
>maintenance makes a better technician as they have already mastered the
>fine arts of duct-tape application, big hammer adjustment techniques,
>baling wire support construction, creative part substitution, and
>applied profanity. The overhead is also less as they've mastered the
>conversion of any convenient tool into a hammer.
I picked a bad car to start learning how to car-work on. Then again, most
any car over the last 15 years is a "bad car" to learn on.
The car still will make it to the dealer at least once a year or so for
software upgrades.
>I've often wondered why the cube-farm or desk pilot variety of
>tech-support is so frustrating. It's been many years since I did this
>kind of work. My guess is that it's like giving flying lessons without
>ever being allowed to fly. It must be terribly frustrating.
[...]
It's more like knowing how to fly quite well, but having to spend your
entire life trying to tell chimpanzees how to land airplanes they've
suddenly become pilots of.
I'm going to bed. Been a pissy day. -rt
Justin The Cynical
[snip]
->I agree. You can never go wrong by using the Scotty method of time estimation.
->Allow enough time to get a quick game of Quake (or whatever) er, I mean, intense
->network testing, in....
Ahh yes. I'm quite familar with this. Every weekend I go to the ISP
to work on the servers[1], and every weekend the best intentions degrade into
a 'network testing' session. Of course, the testing tool happens to be
Quick Understanding API for stacK Enviroments, version 2[2].
Now, if you _really_ want to stress-test a LAN, you need to be using
Dynamic Object Orientated Modules[3], version 1. That is the 'tool' that was
banned from many a company due to the excessive packets being sent onto the
wire.
[1] About five all together. I know the root password on each one. And one
has been decided to be mine. My box, my rules. hehehehe
[2] I can't remember the entire name we had given it, but you get the idea.
[3] More often referred to, in short hand, as DOOM. heh
--
"God, root, what is difference?"
-- Pitr from User Friendly
Justin The Cynical - mo...@ados.com
Highway Star
Liebermann) let loose and wrote:
>You got it easy. My truck just blew a head gasket or cracked the
>head (again). Time to get greasy.
I love my car (see chicken for its name).
It's an '87 with over 150,000 miles on it - and still runs beautifully
(for the most part). The body's kinda battered, thanks to my little
brother who wrecked it at least *3* times (twice in one summer), plus
it has a collection of cop-annoying bumper stickers [1] on the back,
but it's gotten me to places such as Philly, Columbus, and Myrtle
Beach from Charlottesville with no problems but speeding tickets.
I barely have to maintain it...
SeanMike
[1] - For some reason, cops don't like my big "X-COPS" bumper sticker.
Fsck 'em - it's a band, not a statement. Of course, along with all
the "GWAR" ones and the "Fight Crime Shoot Back", I guess it does look
bad. [2]
[2] - Now all I need is the "Bad Cop No Doughnut" sticker.
--
SeanMike Whipkey - all kinds of various stuff here.
I don't speak for my employer unless I say otherwise.
What else do you need to know?
Steve
> Now, if you _really_ want to stress-test a LAN, you need to be using
> Dynamic Object Orientated Modules[3], version 1. That is the 'tool' that was
> banned from many a company due to the excessive packets being sent onto the
> wire.
Around here the Digital Unix disK Extension, ver. 3d, is also quite popular.
One of the few Unix administration tools that runs in DOS. Funny, that...
Scott Brown
Liebermann) wrote:
>On 15 Dec 1998 20:52:41 GMT, rtucker+f...@katan.ttgcitn.com (Ryan
>Tucker) wrote:
>
>Warning: Topic drift.
>
>>I decided it was time to rotate the tires on the Neon, given that my front
>>ones are nearing retirement and my back ones are good enough to pass as
>>new.
>
>I don't believe it.
...
> Nobody does backups.
Sure they do. Usually right after they finish reinstalling the system
after it crashed and burned, because the freshest backups are nineteen
months old and were made right after reinstalling the system after the
*last* crash...
>>night... we'll see what happens in the morning. If worse comes to worse,
>>I may have to impact-wrench the tire guy's testicles. -rt
>
>Remind me not to let you anywhere near my trucks. I use some kind of
>anti-sieze goo on everything exept alloy wheels. Incidentally, the guy
>with the impact wrench has instructions that it's better to torque the
>lug nuts on too tight than to risk litigation when the wheel falls off.
I was a tire monkey in a previous life. You probably don't want to
know the details. These days if I'm having tire work done, I not only
insist that the lugs are hand-tightened, but I bring along my own
torque wrench.
I don't know that I'd want to put anti-sieze compound on lug nuts,
though. A certain amount of siezure is all that's keeping the little
buggers from working their way loose...
>When I was playing RF (radio) engineer, my standard question to
>prospective technicians was "do you work on your own car"? If yes, then
>I would ask for a few details. Generally, anyone that does their own
>maintenance makes a better technician as they have already mastered the
>fine arts of duct-tape application, big hammer adjustment techniques,
>baling wire support construction, creative part substitution, and
>applied profanity. The overhead is also less as they've mastered the
>conversion of any convenient tool into a hammer.
Mmm...I used a 160Mb hard drive as a hammer once. We're still
debating about whether to ship it back to IBM for a warranty
replacement (the drive was kaput before I ever got ahold of it).
Ryan Tucker
<s...@xmission.removethis.com> spewed:
>I was a tire monkey in a previous life. You probably don't want to
>know the details. These days if I'm having tire work done, I not only
>insist that the lugs are hand-tightened, but I bring along my own
>torque wrench.
I'm half-thinking of printing out neat little "HAND TIGHTEN ONLY -- DO NOT
USE IMPACT WRENCH" stickers and sticking them near the lug nuts. The area
is kept clean and dry, so they'd likely survive.
But I drove my car around town a bit after my car's first hand-torquing
experience. Strange, the nasty steering wheel shake at 73mph is almost
completely gone.
I could probably get it completely gone with some more work (they're all
torqued to 90 ft-lbs... plus/minus 5 ft-lbs). But for the first time
using a torque wrench, I think things are damned impressive.
'course, the whole while my dad was mentioning his racing days, back when
they just tightened the nuts to an arbitrary snugness, usually determined
by which note in "Louie Louie" the specific nut was tightened to.
Don't they replace tires every 100 miles or so?
>I don't know that I'd want to put anti-sieze compound on lug nuts,
>though. A certain amount of siezure is all that's keeping the little
>buggers from working their way loose...
An excessive amount of siezure is all that's keeping the little buggers
from getting worked loose when you have to take the motherfuckers off
after an Iowa winter.
I just gave the posts a gentle squirt of WD-40... hey, applied liberally
and left overnight, it loosened The Screw That Not Even God Can Unscrew
enough to be removed with merely 15 minutes of work.
'sides, I'm planning on retorquing every week or two (e.g. every few
thousand miles, if I get stuck doing work on the other side of town. I
gotta get a real job.)... any lack-of-siezure problems will be noted and
fixed with due haste. -rt
Justin The Cynical
<je...@comix.santa-cruz.ca.us> wrote:
->On Tue, 15 Dec 1998 18:03:35 GMT, t...@leaphome.pelican.org (Tim
->Franklin) wrote:
->
->>Isn't SIGTERM supposed to cause things to exit in as polite a fashion as
->>possible? If not, just what *is* its function when we have SIGKILL?
->
->Oh swell. A tech support question in a tech-support newsgroup. What is
->this newsgroup coming to? Support questions are everywhere. I made the
AFAICT, it's following along with the FAQ, as it should be.
Depensing of clues is not prohibited to fellow Bob's. The trick, you see, is
retaining those clues.
->mistake of wearing a Cruzio (a local ISP) T-shirt to a street fair. I
->got stopped a dozen times with "You must work for Cruzio and I have a
->computer problem". I should be like the smarter users and play dumb.
->It attracts fewer support questions.
See the thread a bit ago about the worst places one has ever been
asked a tech support question. The two I recall was at one Bob(ette)'s
wedding, and the other was during a visit to the proctologist(sp?), from the
doctor no less.
[snip UI of SIGKILL, which is allowed by the FAQ. This isn't ASR, ya know.]
Jeff Liebermann
wrote:
> AFAICT, it's following along with the FAQ, as it should be.
>Depensing of clues is not prohibited to fellow Bob's. The trick, you see, is
>retaining those clues.
I leave the camoflage of information to the tech writers. Also, in this
business, there are no answers, only clues. I learned that trick from
the legal profession where an attorney never never never supplies
information, but only offers clues or opinions (same thing). Actually,
an attorney never supplies a single clue because that would presume that
they would actually be responsible for their guesswork. Two or more
opinions|clues|guesses are always supplied as this leaves the
responsibility with the client.
>->mistake of wearing a Cruzio (a local ISP) T-shirt to a street fair. I
>->got stopped a dozen times with "You must work for Cruzio and I have a
>->computer problem". I should be like the smarter users and play dumb.
>->It attracts fewer support questions.
>
> See the thread a bit ago about the worst places one has ever been
>asked a tech support question. The two I recall was at one Bob(ette)'s
>wedding, and the other was during a visit to the proctologist(sp?), from the
>doctor no less.
Sheesh. Getting asked computer questions is so common that I don't even
notice any more. Thinking back this week:
1. I blew the head gasket on my truck. I got a deal from the parts
place in trade for coercing Windoze 95 into saving his passwords.
2. When I asked for advice at an auto shop (one of my customers), it
was strictly quid pro quo and I ended up doing a bit of cleanup on his
file server and tweaking fonts on his printer.
3. The radio club brunch turned into an impromptu lecture on Y2K issues
in the parking lot for 2 hours.
4. My visit to the body mechanic's ended up with a discussion on
UPS/SPS battery backup systems as a the 10 minute UPS didn't quite last
through the last 4 hour power failure.
5. Standing in front of a local ISP's office, waiting for a friend, for
lunch, resulted in customers bugging me with their questions before
blundering into the offices to do the same to the staff.
That's just the last 5 days or so. Next week starts the sacred holiday
of Chistmas where once again I test my survival abilities by living
solely on hors d'oevres. At each of the parties I crash, I'm certain to
get interrogated about computers for the relatives, upgrades, Y2K, etc.
However, this year may be different. Since fearless leader found it
expedient to get a little bombing in before impeachment and Ramadan (a
guaranteed cease-fire), the favorite topic of discussion may finally
drift away from computers.
Scott Brown
(Ryan Tucker) wrote:
>On Thu, 17 Dec 1998 00:37:15 GMT, Scott Brown
><s...@xmission.removethis.com> spewed:
>>I was a tire monkey in a previous life. You probably don't want to
>>know the details. These days if I'm having tire work done, I not only
>>insist that the lugs are hand-tightened, but I bring along my own
>>torque wrench.
>
>I'm half-thinking of printing out neat little "HAND TIGHTEN ONLY -- DO NOT
>USE IMPACT WRENCH" stickers and sticking them near the lug nuts. The area
>is kept clean and dry, so they'd likely survive.
Tire monkeys can't read. Those that can are probably BTMFHs, and are
likely as not to use some JB Weld to "secure" those lug nuts for you.
I would.
>>I don't know that I'd want to put anti-sieze compound on lug nuts,
>>though. A certain amount of siezure is all that's keeping the little
>>buggers from working their way loose...
>
>An excessive amount of siezure is all that's keeping the little buggers
>from getting worked loose when you have to take the motherfuckers off
>after an Iowa winter.
Given the alternative, this is preferable. I've seen one vehicle lose
a tire at highway speeds (very dramatic) and heard the tale of
another. It usually causes lots of *expensive* lower suspension
damage, at the very least.
>I just gave the posts a gentle squirt of WD-40... hey, applied liberally
>and left overnight, it loosened The Screw That Not Even God Can Unscrew
>enough to be removed with merely 15 minutes of work.
I have my own impact wrench for LARTing fastener-type hardware. Best
investment I ever made. Highly recommended.
Alan J Rosenthal
> See the thread a bit ago about the worst places one has ever been
>asked a tech support question. The two I recall was at one Bob(ette)'s
>wedding, and the other was during a visit to the proctologist(sp?), from the
>doctor no less.
"Hey, what I see up your ass reminds me of some Microsoft software I'm having
trouble with."
pachista
je...@comix.santa-cruz.ca.us wrote:
<snip>
>
> That's just the last 5 days or so. Next week starts the sacred holiday
> of Chistmas where once again I test my survival abilities by living
> solely on hors d'oevres. At each of the parties I crash, I'm certain to
> get interrogated about computers for the relatives, upgrades, Y2K, etc.
> However, this year may be different. Since fearless leader found it
> expedient to get a little bombing in before impeachment and Ramadan (a
> guaranteed cease-fire), the favorite topic of discussion may finally
> drift away from computers.
Don't bet on it. They will talk about computers and we will bomb during
Ramadan. (not that I am saying that's a good thing...)
-pachista
--
They tried to give me advice down
at the record shop 'Sit down boys;
this may come as a shock. What's all
I listen to, it's all freedom rock.'- Frank Black
Kenneth Brody
Jeff Liebermann wrote:
[...]
> 3. The server reboots without any error messages or forced fsck.
[...]
and it takes a second or two to remember that "fsck" can also be a Unix
command.
--
+---------+----------------------------------+-----------------------------+
| Kenneth | kenb...@bestweb.net | "The opinions expressed |
| J. | | herein are not necessarily |
| Brody | http://www.bestweb.net/~kenbrody | those of fP Technologies." |
+---------+----------------------------------+-----------------------------+
GCS (ver 3.12) d- s+++: a C++$(+++) ULAVHSC^++++$ P+>+++ L+(++) E-(---)
W++ N+ o+ K(---) w@ M@ V- PS++(+) PE@ Y+ PGP-(+) t+ R@ tv+() b+
DI+(++++) D---() G e* h---- r+++ y?