Line conditioning

[Rick Moen]

In recounting the inherent problems of UPSes (modern 'line-interactive'
ones reportedly prevalent today for modest scenarios needing less than 1kVA
of feed), I mentioned the added complexity of first autotransforming the AC
to constrain its range, then rectifying to DC and feeding that into a
battery, then taking DC off the battery, and running it through an inverter
again. I mentioned the significant power consumption and heat-generation
within the UPS (ironic when one of one's aims is to go greener with home
energy consumption and lower one's power bill). I mention the horror
stories of UPSes melting down and in some cases catching fire.

But there's an additional bit I should also add, which I alluded to briefly
upthread in passing:

People who think a 'UPS' is a magic solution to power problems really
haven't thought things through, and have never had to deal with the
problems caused by failing batteries every few years.

With a UPS, you have a new obligation: Every six months to a year, you have
to recalibrate the UPS's battery-state sensor. These days, the UPS units
do some automated testing of remaining battery quality, and reportedly turn
on a flashing indicator light and start emitting a beep when the periodic
self-test reveals that it thinks you need a new battery pack. Battery packs
last maybe three years in use.

The battery pack is the overwhelming share of the UPS's cost, so you end up
buying 70% of a UPS every few years, or 100% if you follow the path of least
resistance and get a new UPS. (Replacement gel-cel lead-acid batteries are
available from reputable third parties for less than you can get them from
APC, TrippLite, etc.) And, of course, you also need to dispose properly of
the old battery pack, as it's hazardous waste.

That is all the _best-case_ scenario. If the unit's automated self-testing
fails to notify you of a failing battery pack, or you ignore the alerts,
then the UPS eventually ceases ot provide power at all and becomes a single
point of failure for everything bheind it. The failure mode of UPS
batteries is to turn off everything downstream, and there is no automated
fallback to regular AC power. In that case, the thing you bought to
protecdt you from power problems _becomes_ a power problem.

That's the sort of thing I meanmt by the remark that complexity is the enemy
of reliability.

[Scott DuBois]

Post by Rick Moen
In recounting the inherent problems of UPSes (modern 'line-interactive'
ones reportedly prevalent today for modest scenarios needing less than 1kVA
of feed), I mentioned the added complexity of first autotransforming the AC
to constrain its range, then rectifying to DC and feeding that into a
battery, then taking DC off the battery, and running it through an inverter
again. I mentioned the significant power consumption and heat-generation
within the UPS (ironic when one of one's aims is to go greener with home
energy consumption and lower one's power bill). I mention the horror
stories of UPSes melting down and in some cases catching fire.
But there's an additional bit I should also add, which I alluded to briefly
People who think a 'UPS' is a magic solution to power problems really
haven't thought things through, and have never had to deal with the
problems caused by failing batteries every few years.
With a UPS, you have a new obligation: Every six months to a year, you have
to recalibrate the UPS's battery-state sensor. These days, the UPS units
do some automated testing of remaining battery quality, and reportedly turn
on a flashing indicator light and start emitting a beep when the periodic
self-test reveals that it thinks you need a new battery pack. Battery packs
last maybe three years in use.
The battery pack is the overwhelming share of the UPS's cost, so you end up
buying 70% of a UPS every few years, or 100% if you follow the path of least
resistance and get a new UPS. (Replacement gel-cel lead-acid batteries are
available from reputable third parties for less than you can get them from
APC, TrippLite, etc.) And, of course, you also need to dispose properly of
the old battery pack, as it's hazardous waste.
That is all the _best-case_ scenario. If the unit's automated self-testing
fails to notify you of a failing battery pack, or you ignore the alerts,
then the UPS eventually ceases ot provide power at all and becomes a single
point of failure for everything bheind it. The failure mode of UPS
batteries is to turn off everything downstream, and there is no automated
fallback to regular AC power. In that case, the thing you bought to
protecdt you from power problems _becomes_ a power problem.
That's the sort of thing I meanmt by the remark that complexity is the enemy
of reliability.

Yes, it all sounds like a real PITA but what's a person to do? It's the
same as a maintenance free car - it just doesn't happen. All of this is
simply reiterating why Mike has done so well with Hurricane. There are a
large number of people that just don't want to mess with the
complexities involved with managing their own hardware so they turn to a
colo to do it for them or, others simply turn to IaaS or PaaS offerings.
Running a service from the house has it's own level of complexity with
pro's and con's that have to be dealt with.

[Rick Moen]

Post by Scott DuBois
Yes, it all sounds like a real PITA but what's a person to do? It's the
same as a maintenance free car - it just doesn't happen.

The question I always had about UPSes is the one that is forever arising in
computer technology: Are you solving the right problem? Specifically, are
you solving the right problem in _your use scenario_? Vendors will always
happily take your money for gear that is absolutely wrong for your
situation. LUGs, for their part, tend to worsen that problem, being
chock-full of people who (as I show in the cited SVLUG thread) will
cheerfully ignore your carefully explained requirements and rationale, and
recommend whatever they're familiar with despite utter inappropriateness.
Also, a large chunk of computer people are gadget freaks, who are always
eager to overcomplicate things, because they think mechanism is cool and
honestly believe that adding another layer to meet any perceived need is
presumptively a good idea. (Remind them that avoidable complexity impairs
reliability and security, and they just look confused. Why would another
layer of stuff not be good?)

As I said, at root, a UPS is a big honkin' battery, with some supporting
stuff in front of and behind it. The main function of the battery in
this situation is to bridge your uptime over short and medium power
outages[1], and give you some brownout protection.

Here in the Bay Area, we don't really have brownouts, so the key question
is: Do you really need to bridge uptime over short and medium power
outages? Personally, I know I absolutely don't. Ever since journaled
filesystems became routine on Linux, everything comes right back up like
clockwork when the power returns, and I really can't be arsed about 5
minutes of power-outage downtime once every couple of years.

The UPS incidentally (if it is of the line-interactive type) does some
line conditioning, but that is not what it's really designed to do and not
its core competency. Also, you get to deal with all the ongoing battery
bullshit, including replacing the battery pack completely about every three
years. All of which brings back the perennial question: Are you solving
the right problem in the first place? Maybe, if you need anything at all,
you need something that specifically _omits_ an inline expensive battery
pack, a rectifier, and an inverter. Like a line conditioner.

Post by Scott DuBois
Running a service from the house has its own level of complexity with
pros and cons that have to be dealt with.

Rather more often than you might imagine, the right way to deal with a level
of complexity is to carefully avoid it entirely. Overengineering is an
actively bad thing.

[1] This includes, by extension, giving users the opportunity to save
their work and shut down before battery exhausiion during long power
outages. In my use-case, that doesn't matter, either.

[bloonoise]

I usually have a brown out at least once a day;)

Post by Rick Moen
Here in the Bay Area, we don't really have brownouts,

[Scott DuBois]

Post by bloonoise
I usually have a brown out at least once a day;)

Post by Rick Moen
Here in the Bay Area, we don't really have brownouts,

Cute.

Actually, it wasn't too long ago that PG&E would do rolling "browns" to
cut down on consumption and there's always the idiot DUI's and such that
take out power lines which we never really know when OR where such
incidents will happen. We never "really" know what PG&E is going to do
as they don't often "advertise" their intentions that will affect any
particular grid. Ask residents of San Bruno how reliable PG&E is when it
comes to information, I'm sure the results will be enlightening.

Yes, a UPS is a PITA of added complexity to deal with but just like
anything else I think it comes down to how important is up-time to the
application? Weigh the pros and cons and what is "really" needed for the
application as we are discussing in this thread.

In this case, line conditioning is of interest but battery backup? Not
so much; so why deal with the inherent associated BS for something we
don't need?

[bloonoise]

I dunno living in a newer building in SF I haven't really experienced
anything that
would put me at concern. Hetch Hetchy is pretty reliable as noted.
During the rare
thunderstorm I get paranoid and shut every down. I dunno its more
paranoia or based on
fact or how much fact. Nobody wants to loose or corrupt data. Or have a
spike fry your
hardware. Has anyone ever experienced this or heard of it closely.
I doubt that my surge protections are worth much. And never really know
if they are.
Its a crap shoot. Insurance, backup and prayer seem as reliable as anything.

Post by Scott DuBois

Post by bloonoise
I usually have a brown out at least once a day;)

Post by Rick Moen
Here in the Bay Area, we don't really have brownouts,

Cute.
Actually, it wasn't too long ago that PG&E would do rolling "browns" to
cut down on consumption and there's always the idiot DUI's and such that
take out power lines which we never really know when OR where such
incidents will happen. We never "really" know what PG&E is going to do
as they don't often "advertise" their intentions that will affect any
particular grid. Ask residents of San Bruno how reliable PG&E is when it
comes to information, I'm sure the results will be enlightening.
Yes, a UPS is a PITA of added complexity to deal with but just like
anything else I think it comes down to how important is up-time to the
application? Weigh the pros and cons and what is "really" needed for the
application as we are discussing in this thread.
In this case, line conditioning is of interest but battery backup? Not
so much; so why deal with the inherent associated BS for something we
don't need?
_______________________________________________
BALUG-Talk mailing list
http://lists.balug.org/listinfo.cgi/balug-talk-balug.org

[Scott DuBois]

Post by bloonoise
I dunno living in a newer building in SF I haven't really experienced
anything that
would put me at concern. Hetch Hetchy is pretty reliable as noted.
During the rare
thunderstorm I get paranoid and shut every down. I dunno its more
paranoia or based on
fact or how much fact. Nobody wants to loose or corrupt data. Or have a
spike fry your
hardware. Has anyone ever experienced this or heard of it closely.
I doubt that my surge protections are worth much. And never really know
if they are.
Its a crap shoot. Insurance, backup and prayer seem as reliable as anything.

You're right, it is a crap shoot and we can be as much or little
prepared as our paranoia drives us to be. Luckily, since the incident I
had with using a "Christmas Special" surge protector trash a PS from low
voltage output, I personally have not had issues.

But like you said, insurance-backup-prayer seems to work pretty good. :-)

[Rick Moen]

Post by Scott DuBois
Actually, it wasn't too long ago that PG&E would do rolling "browns" to
cut down on consumption

No, that's not what happened during the summer of 2004 during Governor Gray
Davis's administration. It was rolling _blackouts_, which were deliberate
utility-decided shutoffs of parts of the grid during days when PG&E lacked
the capacity to meet demand. Some members of the public and the press
referred to this effect as 'rolling brownouts', but only because they were
confused about the diffference between brownouts and blackouts.

(What the public didn't know ws that the fuel shortage was carefully
engineered by Enron Corporation through market manipulation, to make a
killing in the spot oil market at Californians' expense.)

That was the summer when, finally, getting tired of coming home to walk my
server though a manual fsck after it'd been down most of a workday because
of a 1 hour PG&E blackout in the morning, I painstakingly converted my
filesystems from ext2 to XFS. (This is before ext3 existed.)

Notably, everyone's home Linux systems suffered downtime that summer
(waiting at fsck prompts for their owners to come home), but did _not_
get fried - because those were blackouts, not brownouts.

Brownouts happen when there is a capacity problem and the utility is
totatlly asleep at the wheel, so that significantly reduced voltage
(like, 50-60 volts) starts being delivered for hours at a time.
(A blip for a second when the utility flips switches to shift capacity
around could be called a de-minimus 'brownout', but is not long enough to
fry motherboards by a long shot.)

Post by Scott DuBois
there's always the idiot DUIs and such that take out power lines...

Now, you're reaching.

Post by Scott DuBois
We never "really" know what PG&E is going to do...

Now, you're reaching.

Post by Scott DuBois
Ask residents of San Bruno...

Now, youre reaching.

Post by Scott DuBois
Weigh the pros and cons...

I did.

[Scott DuBois]

Post by Rick Moen

Post by Scott DuBois
Actually, it wasn't too long ago that PG&E would do rolling "browns" to
cut down on consumption

No, that's not what happened during the summer of 2004 during Governor Gray
Davis's administration. It was rolling _blackouts_, which were deliberate
utility-decided shutoffs of parts of the grid during days when PG&E lacked
the capacity to meet demand. Some members of the public and the press
referred to this effect as 'rolling brownouts', but only because they were
confused about the diffference between brownouts and blackouts.
(What the public didn't know ws that the fuel shortage was carefully
engineered by Enron Corporation through market manipulation, to make a
killing in the spot oil market at Californians' expense.)
That was the summer when, finally, getting tired of coming home to walk my
server though a manual fsck after it'd been down most of a workday because
of a 1 hour PG&E blackout in the morning, I painstakingly converted my
filesystems from ext2 to XFS. (This is before ext3 existed.)
Notably, everyone's home Linux systems suffered downtime that summer
(waiting at fsck prompts for their owners to come home), but did _not_
get fried - because those were blackouts, not brownouts.
Brownouts happen when there is a capacity problem and the utility is
totatlly asleep at the wheel, so that significantly reduced voltage
(like, 50-60 volts) starts being delivered for hours at a time.
(A blip for a second when the utility flips switches to shift capacity
around could be called a de-minimus 'brownout', but is not long enough to
fry motherboards by a long shot.)

Post by Scott DuBois
there's always the idiot DUIs and such that take out power lines...

Now, you're reaching.

Post by Scott DuBois
We never "really" know what PG&E is going to do...

Now, you're reaching.

Post by Scott DuBois
Ask residents of San Bruno...

Now, youre reaching.

Post by Scott DuBois
Weigh the pros and cons...

I did.

LOL! You remember those days. :-) Figured you would having hardware to
deal with. I just remember the news reports of energy shortages and
thinking "How the hell can we be sucking up that much power?" Of course,
now, years later and a bit more educated I get it.

Eh, reaching maybe and aside from the PG&E "cluster" in San Bruno, the
other stuff is true and does happen. Not so much storms around here but
interruptions happen. It just all depends on how paranoid and prepared a
person wants to be.

-

[Rick Moen]

During the rare thunderstorm, I get paranoid and shut every[thing] down.
I dunno [if] it's more paranoia or based on fact, or how much fact.
Nobody wants to lose or corrupt data. Or have a spike fry your hardware.
Has anyone ever experienced this or heard of it closely?`

What's the question? If the question is 'Can power irregularities caused by
nearby lightning strikes directly hitting utility infrastructure damage
consumer electronics?', then the answer is yes, especially in the vast
majority of the Bay Area where power lines have not been put underground.

I think it was in 2005, when my previous home server, an antique VA Research
model 501 dating from 1998, was largely destroyed during a heavy wind and
lightning storm. The motherboard, PSU, and RAM were all dead. The hard
drives were OK. This degree of destruction is rare (not even counting the
fact that lightning storms are a bit rare, our not being in Oklahoma).

If there's a lightning storm and you're worried about computers,
televisions, radios, microwave ovens, and so on, the thorough approach would
be to unplug them, not merely shut them down.

I doubt that my surge protections are worth much.

Ever bother to look into what's in them and how they work? I'm betting no,
and that you paid peanuts for whatever it is you're talking about. In that
case, setting your expectations to 'I paid money for rubbish that I don't
even really understand and have no reason to think is useful' seems a good bet.

Consider doing a little reading to better understand the subject, and then
I'll bet you can make better choices.

(If you rally want to be known here as just 'bloonoise', that's fine.
If not, you might want to set the realname field for your mail software.
There's a young guy who posted to Bay Area Linux mailing lists for about a
year as 'lordsauronthegreat', which was his privilege, but he got a bit more
respect after setting his realname field to Chris Miller. Just sayin'.)

[bloonoise]

There is no question, only point basically.
Yes I unplug my equipment as a precaution during thunderstorms.
But beyond that I've had no reason to worry about it much.
This is after many many years.
The power supply's already acts as a capacitor. I have good ones.
It is major fluctuations we should be addressing.
Yes you can go out and spend hundreds maybe thousands on
UPS systems and or switches. But practicality should always be considered.
As far as my persona goes this is but something like my third post.
I don't see where I've been disrespected or given such.
I met Rick many years ago at the old meetings below the church @ Geary
and Gough many years ago. Beyond that I've just got here and am not
in the habit of instantly giving my name to anyone on the Net.
Nor do I think its appropriate. You yourself wrote a piece known broadly
on how
to post. I respect that. I give respect. Perhaps I haven't gotten it I
don't know.
Maybe my name will pop up in time once I know you better.
In so far as my equipment goes it isn't necessarily cheap at all.
I've had no reason to make a comprehensive study on my power grids.
I'm listening and interested and that is why I made a point.

Post by Rick Moen

During the rare thunderstorm, I get paranoid and shut every[thing] down.
I dunno [if] it's more paranoia or based on fact, or how much fact.
Nobody wants to lose or corrupt data. Or have a spike fry your hardware.
Has anyone ever experienced this or heard of it closely?`

What's the question? If the question is 'Can power irregularities caused by
nearby lightning strikes directly hitting utility infrastructure damage
consumer electronics?', then the answer is yes, especially in the vast
majority of the Bay Area where power lines have not been put underground.
I think it was in 2005, when my previous home server, an antique VA Research
model 501 dating from 1998, was largely destroyed during a heavy wind and
lightning storm. The motherboard, PSU, and RAM were all dead. The hard
drives were OK. This degree of destruction is rare (not even counting the
fact that lightning storms are a bit rare, our not being in Oklahoma).
If there's a lightning storm and you're worried about computers,
televisions, radios, microwave ovens, and so on, the thorough approach would
be to unplug them, not merely shut them down.

I doubt that my surge protections are worth much.

Ever bother to look into what's in them and how they work? I'm betting no,
and that you paid peanuts for whatever it is you're talking about. In that
case, setting your expectations to 'I paid money for rubbish that I don't
even really understand and have no reason to think is useful' seems a good bet.
Consider doing a little reading to better understand the subject, and then
I'll bet you can make better choices.
(If you rally want to be known here as just 'bloonoise', that's fine.
If not, you might want to set the realname field for your mail software.
There's a young guy who posted to Bay Area Linux mailing lists for about a
year as 'lordsauronthegreat', which was his privilege, but he got a bit more
respect after setting his realname field to Chris Miller. Just sayin'.)
_______________________________________________
BALUG-Talk mailing list
http://lists.balug.org/listinfo.cgi/balug-talk-balug.org

[jim]

(can't help myself--I've hobnobbed with
power conditioners.)

AC power is tightly controlled with respect
to frequency and voltage. Abnormalities of
any kind are rare.
"Noise" (high frequency small variations in
voltage) is generally negligible.
Browns are the most frequent abnormality,
usually a brief and smallish drop in voltage.
Browns can be sufficiently severe to have the
affect of a black.
Blacks are rare, usually of long enough time
that no ordinary power supply can sustain
working voltage--the equipment dies, and in
some cases revives after the normal voltage
reappears. Hard drives that suffer repeated
deaths sometimes result in nested lost+found
directories. Blackouts don't damage things
in the usual sense of damage (fried
semiconductors).
Highs (voltages that significantly exceed
the normal AC line peaks for more than a
millisecond or so) are rare. Highs can destroy
circuit components. The usual source of highs
comes from motors in the neighborhood.
Lightning presents a special case of highs--
peaks may exceed 1000 VAC over many
milliseconds. The equipment that provides the
electricity for a building typically arcs with
voltage peaks over 6000V.

To protect against browns is likely a matter
of installing additional capacitors in the device,
not practical for most folks.
To protect against blacks requires an
uninterruptible power supply (UPS), which
necessarily depends on batteries; typically the
UPS will last a few minutes. It's a good idea that
a UPS has some means of notifying the
computer and starts a graceful shutdown.

To protect against noise and gentle highs, the
best protection is that of a LC pi filter--a very
small, high working voltage capacitor followed
by a coil (the bigger the better) followed by
another capacitor.
Better protection adds metal oxide varistors
(MOVs) and transient voltage suppression (TVS)
diodes.
Even better protection additionally adds a
quick-blow fuse.
There is no protection against a nearby
lightning strike other than unplugging the
equipment. (Well, there is equipment that
protects against the massive energy of a
lightning strike, but that stuff costs lots.)

Practically, research a protection device to
verify what kind of stuff it's got inside (coils,
caps, MOVs, TVSs, fuses). Note that most
circuit breakers are too slow to protect against
high-power highs.

Post by Rick Moen

During the rare thunderstorm, I get paranoid and shut every[thing] down.
I dunno [if] it's more paranoia or based on fact, or how much fact.
Nobody wants to lose or corrupt data. Or have a spike fry your hardware.
Has anyone ever experienced this or heard of it closely?`

What's the question? If the question is 'Can power irregularities caused by
nearby lightning strikes directly hitting utility infrastructure damage
consumer electronics?', then the answer is yes, especially in the vast
majority of the Bay Area where power lines have not been put underground.
I think it was in 2005, when my previous home server, an antique VA Research
model 501 dating from 1998, was largely destroyed during a heavy wind and
lightning storm. The motherboard, PSU, and RAM were all dead. The hard
drives were OK. This degree of destruction is rare (not even counting the
fact that lightning storms are a bit rare, our not being in Oklahoma).
If there's a lightning storm and you're worried about computers,
televisions, radios, microwave ovens, and so on, the thorough approach would
be to unplug them, not merely shut them down.

I doubt that my surge protections are worth much.

Ever bother to look into what's in them and how they work? I'm betting no,
and that you paid peanuts for whatever it is you're talking about. In that
case, setting your expectations to 'I paid money for rubbish that I don't
even really understand and have no reason to think is useful' seems a good bet.
Consider doing a little reading to better understand the subject, and then
I'll bet you can make better choices.
(If you rally want to be known here as just 'bloonoise', that's fine.
If not, you might want to set the realname field for your mail software.
There's a young guy who posted to Bay Area Linux mailing lists for about a
year as 'lordsauronthegreat', which was his privilege, but he got a bit more
respect after setting his realname field to Chris Miller. Just sayin'.)
_______________________________________________
BALUG-Talk mailing list
http://lists.balug.org/listinfo.cgi/balug-talk-balug.org

[Rick Moen]

Beyond that I've just got here and am not in the habit of instantly giving
my name to anyone on the Net. Nor do I think its appropriate.

As I said, entirely your right, just as it was lordsauronthegreat's.
I mentioned it only because many folks fail to realise they've _forgotten_
to set a realname field.

Heck, we have a guy around here who posts as Maestro, and if you encounter
him in person, he introduces himself as Maestro. Works for him, apparently.
Far as I'm concerned, he could also call himself Liberace. My undestanding
is that said handle's available again after a long monopoly by an
entertainer.

Yes you can go out and spend hundreds maybe thousands on UPS systems and
or switches. But practicality should always be considered.

No kidding? I never thought of using practicality. Is that often on sale
at Fry's? ;->

Yes I unplug my equipment as a precaution during thunderstorms.

As an afterthought to that, it might be worthwhile also temporarily
unplugging any cable to broadband or telephone (modem) connectors for the
same reason.

In so far as my equipment goes it isn't necessarily cheap at all.

You're the one who was susggesting you suspected your surge protectors don't
do any good. So, bloonoise, what'cha got?

I've had no reason to make a comprehensive study on my power grids.
I'm listening and interested and that is why I made a point.

You mean, you want this all explained, but would rather not do any work?
This is indeed a common sentiment. How's it working for you? ;->

[bloonoise]

lol.......
anyways my name is Hardy X esq.
I'm presently drinking wine and watching the Roosevelt's on PBS
so as admin maybe you can set that right; Hardy.
if I do I'm not responsible right now. And I'll screw it up probably.
ttyl.........

Yeah all conductive things come unplugged in storms
fortunately we don't have thunderstorms often , enough, could use
one.........;)

Post by Rick Moen

Beyond that I've just got here and am not in the habit of instantly giving
my name to anyone on the Net. Nor do I think its appropriate.

As I said, entirely your right, just as it was lordsauronthegreat's.
I mentioned it only because many folks fail to realise they've _forgotten_
to set a realname field.
Heck, we have a guy around here who posts as Maestro, and if you encounter
him in person, he introduces himself as Maestro. Works for him, apparently.
Far as I'm concerned, he could also call himself Liberace. My undestanding
is that said handle's available again after a long monopoly by an
entertainer.

Yes you can go out and spend hundreds maybe thousands on UPS systems and
or switches. But practicality should always be considered.

No kidding? I never thought of using practicality. Is that often on sale
at Fry's? ;->

Yes I unplug my equipment as a precaution during thunderstorms.

As an afterthought to that, it might be worthwhile also temporarily
unplugging any cable to broadband or telephone (modem) connectors for the
same reason.

In so far as my equipment goes it isn't necessarily cheap at all.

You're the one who was susggesting you suspected your surge protectors don't
do any good. So, bloonoise, what'cha got?

I've had no reason to make a comprehensive study on my power grids.
I'm listening and interested and that is why I made a point.

You mean, you want this all explained, but would rather not do any work?
This is indeed a common sentiment. How's it working for you? ;->
_______________________________________________
BALUG-Talk mailing list
http://lists.balug.org/listinfo.cgi/balug-talk-balug.org

[Scott DuBois]

Post by Rick Moen

Beyond that I've just got here and am not in the habit of instantly giving
my name to anyone on the Net. Nor do I think its appropriate.

As I said, entirely your right, just as it was lordsauronthegreat's.
I mentioned it only because many folks fail to realise they've _forgotten_
to set a realname field.
Heck, we have a guy around here who posts as Maestro, and if you encounter
him in person, he introduces himself as Maestro. Works for him, apparently.
Far as I'm concerned, he could also call himself Liberace. My undestanding
is that said handle's available again after a long monopoly by an
entertainer.

Yes you can go out and spend hundreds maybe thousands on UPS systems and
or switches. But practicality should always be considered.

No kidding? I never thought of using practicality. Is that often on sale
at Fry's? ;->

Yes I unplug my equipment as a precaution during thunderstorms.

As an afterthought to that, it might be worthwhile also temporarily
unplugging any cable to broadband or telephone (modem) connectors for the
same reason.

In so far as my equipment goes it isn't necessarily cheap at all.

You're the one who was susggesting you suspected your surge protectors don't
do any good. So, bloonoise, what'cha got?

I've had no reason to make a comprehensive study on my power grids.
I'm listening and interested and that is why I made a point.

You mean, you want this all explained, but would rather not do any work?
This is indeed a common sentiment. How's it working for you? ;->
_______________________________________________
BALUG-Talk mailing list
http://lists.balug.org/listinfo.cgi/balug-talk-balug.org

As long as everyone plays nice, I don't care too much. The way I see it,
people cover their identity for two reasons; safety from online creeps
and to be an online creep. So far I see no signs of the latter but yeah,
it becomes "awkward" after a while.

[Rick Moen]

Post by Scott DuBois
Eh, reaching maybe and aside from the PG&E "cluster" in San Bruno, the
other stuff is true and does happen. Not so much storms around here but
interruptions happen.

Interruptions happen, but they merely swtich your machine off. They don't
fry motherboards.

Big spikes, big surges, and brownouts (lengthy and severe undervoltage
conditions) fry motherboards.

In the Bay Area, espeicailly parts thereof where you're not right next door
to a factory or machine shop, we have clean power by any reasonable
standard, and we really don't have brownouts.

Which is what I said in the first place.

Post by Scott DuBois
It just all depends on how paranoid and prepared a person wants to be.

Paranoia's cheap. I'm trying to provide reasonable perspective and
pragmatic advice based on very long experience. For some reason, you're
arguing when I attempt to provide it -- more than just provide it; give it
away for free, pro bono publico -- with you inventing astonishing
objectiions like imaginary brownouts and the San Bruno gas explosion.

It's an Internet thing, of course. As I explained to a reporter many years
ago, all I have to do on the Internet is say 'I like herring', and doing so
will magically summon every dedicated herring-hater for hundreds of miles
around to try to argue that I'm wrong to like herring.

[Scott DuBois]

Post by Rick Moen

Post by Scott DuBois
Eh, reaching maybe and aside from the PG&E "cluster" in San Bruno, the
other stuff is true and does happen. Not so much storms around here but
interruptions happen.

Interruptions happen, but they merely swtich your machine off. They don't
fry motherboards.
Big spikes, big surges, and brownouts (lengthy and severe undervoltage
conditions) fry motherboards.
In the Bay Area, espeicailly parts thereof where you're not right next door
to a factory or machine shop, we have clean power by any reasonable
standard, and we really don't have brownouts.
Which is what I said in the first place.

Post by Scott DuBois
It just all depends on how paranoid and prepared a person wants to be.

Paranoia's cheap. I'm trying to provide reasonable perspective and
pragmatic advice based on very long experience. For some reason, you're
arguing when I attempt to provide it -- more than just provide it; give it
away for free, pro bono publico -- with you inventing astonishing
objectiions like imaginary brownouts and the San Bruno gas explosion.
It's an Internet thing, of course. As I explained to a reporter many years
ago, all I have to do on the Internet is say 'I like herring', and doing so
will magically summon every dedicated herring-hater for hundreds of miles
around to try to argue that I'm wrong to like herring.

I actually didn't see any argument really, my apologies if the
discussion was perceived that way.

I have my personal perspective on certain public utilities based on
criteria which others may or may not find valid reasoning and that's
fine. It's my personal reasoning and my perspective of which I do not
necessarily expect others to share or rational; that's ok. However, the
reverse is also true where others would disagree and have their
reasoning to validate their perspective to which they have every right
to do. I however, choose whether to accept that reasoning or not; that
is my prerogative.

[Rick Moen]

Post by Scott DuBois
As long as everyone plays nice, I don't care too much. The way I see it,
people cover their identity for two reasons; safety from online creeps
and to be an online creep. So far I see no signs of the latter but yeah,
it becomes "awkward" after a while.

I fought long and hard, back in the day, for the legal and practical right
to be pseudonymous on the Internet. There is a traditional laundry list of
outright-noble uses of them that, and ones to which anyone would be
sympathetic like 'avoiding stalkers', always get trotted out by the
lordsauronthegreats of the world when people look askance. As it turns out,
though, the overwhelming use to which pseudonymity gets put is to avoid any
likelhood of accountability. This aim gets dressed up in various
euphemisims like 'I want to be free to speak my mind without any risk of
getting in trouble with my employer', but it basically boils down to
avoiding accountability.

Internet attempts at character assassination, and a large variety of other
misbehaviour, overwhelmingly get launched from psuedonymous throwaway
identities.

Every time I talk to Maestro (though, technically, he goes by 'maestro',
like he feels he's not worthy of proper nouns), I keep expecting that he's
going to pull out a baton and start conducting some invisible orchestra.

[Scott DuBois]

Post by Rick Moen

Post by Scott DuBois
As long as everyone plays nice, I don't care too much. The way I see it,
people cover their identity for two reasons; safety from online creeps
and to be an online creep. So far I see no signs of the latter but yeah,
it becomes "awkward" after a while.

I fought long and hard, back in the day, for the legal and practical right
to be pseudonymous on the Internet. There is a traditional laundry list of
outright-noble uses of them that, and ones to which anyone would be
sympathetic like 'avoiding stalkers', always get trotted out by the
lordsauronthegreats of the world when people look askance. As it turns out,
though, the overwhelming use to which pseudonymity gets put is to avoid any
likelhood of accountability. This aim gets dressed up in various
euphemisims like 'I want to be free to speak my mind without any risk of
getting in trouble with my employer', but it basically boils down to
avoiding accountability.
Internet attempts at character assassination, and a large variety of other
misbehaviour, overwhelmingly get launched from psuedonymous throwaway
identities.

I understand, and agree it goes both ways and unfortunately the Internet
culture has largely adopted "psuedonymous throwaway
identities" for malicious purposes adopted by people with "malicious
intent" thus making it difficult for persons to adopt said psuedonymous
identities for safety and security reasons without being profiled as
having "malicious intent" or, lack of accountability.

Although, creepy, stalker employers these days make it difficult to have
open conversations with others without either hiding one's identity or
having to tack some legal disclaimer on everything they say stating that
"their opinion does not reflect the opinion of their employer". In such
cases, it almost seems people can't be themselves any more as they
become 24/7 company property and representatives.

(I think we covered this not too long ago)??

Post by Rick Moen
Every time I talk to Maestro (though, technically, he goes by 'maestro',
like he feels he's not worthy of proper nouns), I keep expecting that he's
going to pull out a baton and start conducting some invisible orchestra.

Good one! ;-)

[Rick Moen]

Jim wrote:

[blackouts:]

Hard drives that suffer repeated deaths sometimes result in nested
lost+found directories.

Jim, you don't seem old enough as a Linux user to predate journaled
filesystems.

Since adopting XFS (and then later ext3) during the Enron summer of 2004,
I've had absolutely nothing placed by fsck in any filesystem's lost+found
directory, and certainly not nested lost+found directories. Not on any
system, and I've administered thousands of Linux systems.

That is exactly what uses of journaling ought to accomplish, absent some
horrible hardware or software failure out of journaling's scope. Your power
comes back on, the partial uncommited transaction gets backed out to the
last checkpointed state in the journal, and the filesystem gets mounted with
nothing at all being written to lost+found. No matter how many times you
lose and regain power, same result.

Mind you, I am not recommending pulling and replugging the power cable a
hundred times in a row to test the efficacy of journaling. That will
stress the electronics a bit every time.

But within reason, I myself feel I have little to fear from occasional
blackouts - because of experience, and understanding the basics of how
filesystem journaling works. Substantial brownouts (like, half an hour
plus) and major spikes and surges, those are a threat if they happen.

Browns are the most frequent abnormality, usually a brief and smallish
drop in voltage.

Which is why I qualified what I said as _substantial_ brownouts being a
threat to gear (if they occur), and that we basically don't get those
around here. The tiny blip-drops as PG&E switching shifts feeds to
accomodate changing load don't hurt anything.

On the blackout thing, I'm willing to do a $50 wager: I'll find a throwaway
PIII and put a throwaway hard drive in it. We put a single ext3 root
filesystem on it (default journaling and other options), and install
$SOME_DISTRO. We verify that /lost+found is empty. Then, we yank power,
wait 15 seconds, restart Linux host to fully running state. Repeat yank and
restart cycle nine more times, for a total of ten. If there's anything in
/lost+found at the end, I pay you $50. If there isn't, you pay me $50.

If the machine hardware fails, nobody pays, as this is a test of the effect
of journaling, not of old hardware's ability to withstand repeated
power-cycing stress. (I have faith in the work of Ted T'so on ext3, as it's
battle-tested.)

[jim]

If you move the decimal to the left one place and
speed up the repower time to three seconds, I'll
take the bet, mainly for the fun of it.
I haven't anything like the amount of experience
you've got, but just a few years ago I worked on a
drive that had suffered power up-downs and had a
nested lost+found tree. The only explanation I can
figure is that things got horked repeatedly and
sufficiently frequently. I forget which filesystem
was used, but pretty sure it was one of the ext?
systems.

My posting was somewhat indulgent, I know a
little about that stuff and just had to emit--I tried
to write a summary explanation that might help
people who knew less than I to get a useful
perspective.

Post by Rick Moen
[blackouts:]

Hard drives that suffer repeated deaths sometimes result in nested
lost+found directories.

Jim, you don't seem old enough as a Linux user to predate journaled
filesystems.
Since adopting XFS (and then later ext3) during the Enron summer of 2004,
I've had absolutely nothing placed by fsck in any filesystem's lost+found
directory, and certainly not nested lost+found directories. Not on any
system, and I've administered thousands of Linux systems.
That is exactly what uses of journaling ought to accomplish, absent some
horrible hardware or software failure out of journaling's scope. Your power
comes back on, the partial uncommited transaction gets backed out to the
last checkpointed state in the journal, and the filesystem gets mounted with
nothing at all being written to lost+found. No matter how many times you
lose and regain power, same result.
Mind you, I am not recommending pulling and replugging the power cable a
hundred times in a row to test the efficacy of journaling. That will
stress the electronics a bit every time.
But within reason, I myself feel I have little to fear from occasional
blackouts - because of experience, and understanding the basics of how
filesystem journaling works. Substantial brownouts (like, half an hour
plus) and major spikes and surges, those are a threat if they happen.

Browns are the most frequent abnormality, usually a brief and smallish
drop in voltage.

Which is why I qualified what I said as _substantial_ brownouts being a
threat to gear (if they occur), and that we basically don't get those
around here. The tiny blip-drops as PG&E switching shifts feeds to
accomodate changing load don't hurt anything.
On the blackout thing, I'm willing to do a $50 wager: I'll find a throwaway
PIII and put a throwaway hard drive in it. We put a single ext3 root
filesystem on it (default journaling and other options), and install
$SOME_DISTRO. We verify that /lost+found is empty. Then, we yank power,
wait 15 seconds, restart Linux host to fully running state. Repeat yank and
restart cycle nine more times, for a total of ten. If there's anything in
/lost+found at the end, I pay you $50. If there isn't, you pay me $50.
If the machine hardware fails, nobody pays, as this is a test of the effect
of journaling, not of old hardware's ability to withstand repeated
power-cycing stress. (I have faith in the work of Ted T'so on ext3, as it's
battle-tested.)
_______________________________________________
BALUG-Talk mailing list
http://lists.balug.org/listinfo.cgi/balug-talk-balug.org

[Rick Moen]

If you move the decimal to the left one place and speed up the repower
time to three seconds, I'll take the bet, mainly for the fun of it.

What does 'move the decimal to the left one place' mean, 100 repeats
instead of 10? A shirtsleeve estimate suggests that we'd be there about
half a day instead of about an hour. Sorry, but I don't covet your $50
that much, and have better things to do with my time.

That phrase cannot apply to my suggestion of a 15 second delay between
each startup to full runtime mode and pulling the power, because 15
seconds times 0.1 does not equal 3 seconds.

The reason I said 15 seconds was that I don't especially care to even
break a PIII and a spare hard drive, just to prove a point. Now, if you
wish to _purchase_ such a system from me to run this test on, I will be
fine with 3 seconds.

However, 100 repetitions rather than 10 is excessive: Not only would
that consume an entire afternoon, but also, if it were correct that
repeated blackouts create not only chunks of files in lost+found but
also sometimes _nested_ copies of lost+found (something I've never
heard of, before), then IMO this effect really ought to show up during
ten harsh and rapid power-loss-and-reset cycles. (I'm _not_ at all
saying something like what you describe didn't happen, just that I find
zero other examples of it elsewhere.) If you're talking about something
that might appear if we spend an entire afternoon looking for it, during
which we simulate a quantity of power outages that otherwise would
require about thirty years, then I think you're talking about an effect
characterised by Nessie-like levels of 'shyness' from public study.

I haven't anything like the amount of experience you've got, but just
a few years ago I worked on a drive that had suffered power up-downs
and had a nested lost+found tree. The only explanation I can figure is
that things got horked repeatedly and sufficiently frequently.

I'll just point out that your data doesn't actually correlate the one
with the other, even on a sample-of-one basis.

Also and more important, fsck never creates a lost+found subtree within
an existing one. It just doesn't: The code doesn't include any routine
that would do that. So, whatever created such a bizarre situation, it
wasn't fsck -- which is also the only code that's supposed to write there.

As a side note, back in ext2 days, when you _did_ see fsck occasionally
put chunks into lost+found following a loss of power or kernel panic,
in every case _I_ ever experienced, it turned out that these chunks were
redundant to real, non-lost files, such that you then ended up just
deleting them after reviewing what they were. All fsck utilities make
an educated guess about how to correct filesystem state, but fsck.ext2
(aka e2fsck) was extremely reliable about guessing right, and created
those chunks merely out of conservatism in case you needed to
second-guess its corrections. I never saw a case where the chunk wasn't
redundant.

fsck.ext2 was merely (1) nerve-wracking because you weren't _sure_ its
corrections were right until you had checked, and (2) annoying in that
(out of conservatism) if any but the very most trivial fixes were
required upon reboot, fsck.ext2 bailed on the automatic correction
process and exited to a root-console fsck prompt. Thus my extreme
annoyance at Enron Corporation causing my server to sit all workday long
at a manual fsck prompt instead of proceeding through fsck and returning
to operation.

I know a little about that stuff and just had to emit--I tried to
write a summary explanation that might help people who knew less than
I to get a useful perspective.

I do appreciate that, Jim. I just politely differ with your
characterisation of blackouts as dangerous to (journaled) filesystems.
That's just never been the case in my experience. The power comes back
up, fsck backs out any unfinished writes to the last journal checkpoint,
and you're back up witout damage, every time like clockwork (absent
other problems entirely like hardware overstressing).

I'm making that point not on account of being a paid lobbyist for
blackouts, but rather because I'm trying to help readers know what
threats are worth worrying about. Telling them to fear that blackouts
will damage their filesystems in the era of journaled filesystems is
just not correct, and doing readers no favour. That's really my point,
that there are real threats, but this isn't among them. (I don't
actually covet your greenish rectangle w/picture of U.S. Grant. ;->)

(BTW, satellite uplinks suck mightily for ssh. No huge surprise there.)

[jim]

$50.00 ---> $5.00
I don't think fast power up-down will physically
hurt the drive. I believe I have a few drives lying
around and if so will be happy to provide.
I didn't mean to suggest 100 sequential power
up-downs: sorry if I did.
I think a total of thirty minutes trying this out
is a good amount of time, many more minutes
would not be. Perhaps a fun side-project on
some Saturday afternoon?

I encountered the nested lost+found one time
in my life, and that host was having power supply
problems (repeated power up-downs), hence my
inference. If this is a rare corner case, then that
worry about blackouts is specious. In any case,
I'm just inferring that repeated power up-downs
were the cause of the nested l+f tree. The only
claim I can make is that the two conditions were
present on the same box coming in for repairs.

Post by Rick Moen

If you move the decimal to the left one place and speed up the repower
time to three seconds, I'll take the bet, mainly for the fun of it.

What does 'move the decimal to the left one place' mean, 100 repeats
instead of 10? A shirtsleeve estimate suggests that we'd be there about
half a day instead of about an hour. Sorry, but I don't covet your $50
that much, and have better things to do with my time.
That phrase cannot apply to my suggestion of a 15 second delay between
each startup to full runtime mode and pulling the power, because 15
seconds times 0.1 does not equal 3 seconds.
The reason I said 15 seconds was that I don't especially care to even
break a PIII and a spare hard drive, just to prove a point. Now, if you
wish to _purchase_ such a system from me to run this test on, I will be
fine with 3 seconds.
However, 100 repetitions rather than 10 is excessive: Not only would
that consume an entire afternoon, but also, if it were correct that
repeated blackouts create not only chunks of files in lost+found but
also sometimes _nested_ copies of lost+found (something I've never
heard of, before), then IMO this effect really ought to show up during
ten harsh and rapid power-loss-and-reset cycles. (I'm _not_ at all
saying something like what you describe didn't happen, just that I find
zero other examples of it elsewhere.) If you're talking about something
that might appear if we spend an entire afternoon looking for it, during
which we simulate a quantity of power outages that otherwise would
require about thirty years, then I think you're talking about an effect
characterised by Nessie-like levels of 'shyness' from public study.

I haven't anything like the amount of experience you've got, but just
a few years ago I worked on a drive that had suffered power up-downs
and had a nested lost+found tree. The only explanation I can figure is
that things got horked repeatedly and sufficiently frequently.

I'll just point out that your data doesn't actually correlate the one
with the other, even on a sample-of-one basis.
Also and more important, fsck never creates a lost+found subtree within
an existing one. It just doesn't: The code doesn't include any routine
that would do that. So, whatever created such a bizarre situation, it
wasn't fsck -- which is also the only code that's supposed to write there.
As a side note, back in ext2 days, when you _did_ see fsck occasionally
put chunks into lost+found following a loss of power or kernel panic,
in every case _I_ ever experienced, it turned out that these chunks were
redundant to real, non-lost files, such that you then ended up just
deleting them after reviewing what they were. All fsck utilities make
an educated guess about how to correct filesystem state, but fsck.ext2
(aka e2fsck) was extremely reliable about guessing right, and created
those chunks merely out of conservatism in case you needed to
second-guess its corrections. I never saw a case where the chunk wasn't
redundant.
fsck.ext2 was merely (1) nerve-wracking because you weren't _sure_ its
corrections were right until you had checked, and (2) annoying in that
(out of conservatism) if any but the very most trivial fixes were
required upon reboot, fsck.ext2 bailed on the automatic correction
process and exited to a root-console fsck prompt. Thus my extreme
annoyance at Enron Corporation causing my server to sit all workday long
at a manual fsck prompt instead of proceeding through fsck and returning
to operation.

I know a little about that stuff and just had to emit--I tried to
write a summary explanation that might help people who knew less than
I to get a useful perspective.

I do appreciate that, Jim. I just politely differ with your
characterisation of blackouts as dangerous to (journaled) filesystems.
That's just never been the case in my experience. The power comes back
up, fsck backs out any unfinished writes to the last journal checkpoint,
and you're back up witout damage, every time like clockwork (absent
other problems entirely like hardware overstressing).
I'm making that point not on account of being a paid lobbyist for
blackouts, but rather because I'm trying to help readers know what
threats are worth worrying about. Telling them to fear that blackouts
will damage their filesystems in the era of journaled filesystems is
just not correct, and doing readers no favour. That's really my point,
that there are real threats, but this isn't among them. (I don't
actually covet your greenish rectangle w/picture of U.S. Grant. ;->)
(BTW, satellite uplinks suck mightily for ssh. No huge surprise there.)
_______________________________________________
BALUG-Talk mailing list
http://lists.balug.org/listinfo.cgi/balug-talk-balug.org

[bloonoise]

Yeah as mentioned a good suggestion is to upgrade your power supply in
your box.
I've also had problems with the cheap ones that come with the machines.
They
can be problematic. But a simple upgrade solved all problems I was
having. Power
supply's are pretty dam cheap on the net. Like 25 bucks for 450 watts.
Was certainly
worth it for me. I'm running an SSD on my main machine. Dunno how
sensitive SSD's
might be to fluctuations.
SSD's are another upgrade well worth the investment. I got a Samsung 128
pro.
And what a difference! Now boot times are minimal, very. In fact its
made me a bit lazy
Whenever I run into a problem or crash I just kill X and back in in
seconds. So I end up
not tracing or debugging things. Not sure if thats good or bad.
I think the future holds machines where the system / is on an SSD and
all the Data on
one or two or more HD's. I've been very pleased with my SSD. I recommend
them over
fancy surge and UPS devices as an upgrade. Unless you know you need one
or have had
problems. I think UPS devices are more appropriate for servers. I know
all hosting sites use
them there.But they're expensive , and have a finite life cycle....I
dunno doesn't sound feasible
to me for a home Desktop or two or three.
SSD's are great though these days. Coupled with HD's. Samsung pro 128 is
what I've had luck
with. Just gotta mount with Trim enabled and run trim now and
then(weekly cron or manually)..
Supposedly the new ones have life spans equal to HD's.

Post by jim
$50.00 ---> $5.00
I don't think fast power up-down will physically
hurt the drive. I believe I have a few drives lying
around and if so will be happy to provide.
I didn't mean to suggest 100 sequential power
up-downs: sorry if I did.
I think a total of thirty minutes trying this out
is a good amount of time, many more minutes
would not be. Perhaps a fun side-project on
some Saturday afternoon?
I encountered the nested lost+found one time
in my life, and that host was having power supply
problems (repeated power up-downs), hence my
inference. If this is a rare corner case, then that
worry about blackouts is specious. In any case,
I'm just inferring that repeated power up-downs
were the cause of the nested l+f tree. The only
claim I can make is that the two conditions were
present on the same box coming in for repairs.

Post by Rick Moen

If you move the decimal to the left one place and speed up the repower
time to three seconds, I'll take the bet, mainly for the fun of it.

What does 'move the decimal to the left one place' mean, 100 repeats
instead of 10? A shirtsleeve estimate suggests that we'd be there about
half a day instead of about an hour. Sorry, but I don't covet your $50
that much, and have better things to do with my time.
That phrase cannot apply to my suggestion of a 15 second delay between
each startup to full runtime mode and pulling the power, because 15
seconds times 0.1 does not equal 3 seconds.
The reason I said 15 seconds was that I don't especially care to even
break a PIII and a spare hard drive, just to prove a point. Now, if you
wish to _purchase_ such a system from me to run this test on, I will be
fine with 3 seconds.
However, 100 repetitions rather than 10 is excessive: Not only would
that consume an entire afternoon, but also, if it were correct that
repeated blackouts create not only chunks of files in lost+found but
also sometimes _nested_ copies of lost+found (something I've never
heard of, before), then IMO this effect really ought to show up during
ten harsh and rapid power-loss-and-reset cycles. (I'm _not_ at all
saying something like what you describe didn't happen, just that I find
zero other examples of it elsewhere.) If you're talking about something
that might appear if we spend an entire afternoon looking for it, during
which we simulate a quantity of power outages that otherwise would
require about thirty years, then I think you're talking about an effect
characterised by Nessie-like levels of 'shyness' from public study.

I haven't anything like the amount of experience you've got, but just
a few years ago I worked on a drive that had suffered power up-downs
and had a nested lost+found tree. The only explanation I can figure is
that things got horked repeatedly and sufficiently frequently.

I'll just point out that your data doesn't actually correlate the one
with the other, even on a sample-of-one basis.
Also and more important, fsck never creates a lost+found subtree within
an existing one. It just doesn't: The code doesn't include any routine
that would do that. So, whatever created such a bizarre situation, it
wasn't fsck -- which is also the only code that's supposed to write there.
As a side note, back in ext2 days, when you _did_ see fsck occasionally
put chunks into lost+found following a loss of power or kernel panic,
in every case _I_ ever experienced, it turned out that these chunks were
redundant to real, non-lost files, such that you then ended up just
deleting them after reviewing what they were. All fsck utilities make
an educated guess about how to correct filesystem state, but fsck.ext2
(aka e2fsck) was extremely reliable about guessing right, and created
those chunks merely out of conservatism in case you needed to
second-guess its corrections. I never saw a case where the chunk wasn't
redundant.
fsck.ext2 was merely (1) nerve-wracking because you weren't _sure_ its
corrections were right until you had checked, and (2) annoying in that
(out of conservatism) if any but the very most trivial fixes were
required upon reboot, fsck.ext2 bailed on the automatic correction
process and exited to a root-console fsck prompt. Thus my extreme
annoyance at Enron Corporation causing my server to sit all workday long
at a manual fsck prompt instead of proceeding through fsck and returning
to operation.

I know a little about that stuff and just had to emit--I tried to
write a summary explanation that might help people who knew less than
I to get a useful perspective.

I do appreciate that, Jim. I just politely differ with your
characterisation of blackouts as dangerous to (journaled) filesystems.
That's just never been the case in my experience. The power comes back
up, fsck backs out any unfinished writes to the last journal checkpoint,
and you're back up witout damage, every time like clockwork (absent
other problems entirely like hardware overstressing).
I'm making that point not on account of being a paid lobbyist for
blackouts, but rather because I'm trying to help readers know what
threats are worth worrying about. Telling them to fear that blackouts
will damage their filesystems in the era of journaled filesystems is
just not correct, and doing readers no favour. That's really my point,
that there are real threats, but this isn't among them. (I don't
actually covet your greenish rectangle w/picture of U.S. Grant. ;->)
(BTW, satellite uplinks suck mightily for ssh. No huge surprise there.)
_______________________________________________
BALUG-Talk mailing list
http://lists.balug.org/listinfo.cgi/balug-talk-balug.org

_______________________________________________
BALUG-Talk mailing list
http://lists.balug.org/listinfo.cgi/balug-talk-balug.org

[Rick Moen]

Post by Scott DuBois
It's my personal reasoning and my perspective of which I do not
necessarily expect others to share or rational; that's ok. However, the
reverse is also true where others would disagree and have their reasoning
to validate their perspective to which they have every right to do. I
however, choose whether to accept that reasoning or not; that is my
prerogative.

Sure. Fair enough, and well said.

[bloonoise]

My brane just melted and flowed out my left ear canal.......

Post by Rick Moen

Post by Scott DuBois
It's my personal reasoning and my perspective of which I do not
necessarily expect others to share or rational; that's ok. However, the
reverse is also true where others would disagree and have their reasoning
to validate their perspective to which they have every right to do. I
however, choose whether to accept that reasoning or not; that is my
prerogative.

Sure. Fair enough, and well said.
_______________________________________________
BALUG-Talk mailing list
http://lists.balug.org/listinfo.cgi/balug-talk-balug.org

[Scott DuBois]

Post by Rick Moen

Post by Scott DuBois
It's my personal reasoning and my perspective of which I do not
necessarily expect others to share or rational; that's ok. However, the
reverse is also true where others would disagree and have their reasoning
to validate their perspective to which they have every right to do. I
however, choose whether to accept that reasoning or not; that is my
prerogative.

Sure. Fair enough, and well said.

Thank you. Only paraphrasing some of the wisdom passed to me from my
Mom; translated from:

"Everyone has right to their opinion and to voice that opinion, however,
we have the right to listen to or accept that opinion."

and the one that was written on a 3x5 hanging on the refrigerator until
it was yellowed...

"If you want to see past the horizon; you've got to get your head above
the clouds."

[Nick Moffitt]

No Scott, you are not entitled to your opinion.

[bloonoise]

Everyones got an opinion,
the problem is only mine are worth anything at all.

Post by Nick Moffitt
No Scott, you are not entitled to your opinion.

[Rick Moen]

Post by jim
$50.00 ---> $5.00

Oh, OK! Sorry, I didn't get the context. Sure, I like this; makes it more
like the gentleman's wager I intended. (Though, to be honest, it's most
fairly described as a cheap rhetorical device Iuse on rare occasions to
dramatise a point that I feared was getting lost in a messy Internet
discussion.)

Sure, if you want to do that. At a $5 level, I don't even mind if I'm
wrong, as it's a small price for learning something interesting.

Post by jim
I don't think fast power up-down will physically hurt the drive. I believe
I have a few drives lying around and if so will be happy to provide.

Bring the computer as well, and we have a deal. As I said, I think that
rapid a coldboot cycle stresses unduly the hard drive _and_ computer, and
I'm disinclined to throw away a PII just for a wager to make a point.

Post by jim
The only claim I can make is that the two conditions were present on the
same box coming in for repairs.

Put that way, I see your perspective and understand why you reached that
tentative conclusion.

I'm pretty sure it inherently couldn't have been fsck.

fsck creates files (on the increasingly rare cases when it creates anything),
inside a fiiesystem's lost+found directory, with names like '#7479417'.
These are near-invariably _former_ files that fsck temporarily resuscitates,
that were already unlinked (i.e., their names had been erased), but
were still being kept open by some process (so the data wasn't erased yet)
at the moment when the system suddenly halted (kernel panic or power
failure).

fsck does this because there's a very low but non-zero chance that its
guessing algorithm for fixing irregular filesystem structures may make a bad
guess, so it errs on the side of caution in reviving inside lost+found
any data fragment it finds (unlinked) that _might_ possibly be a wanted file
(or portion of such a file).

It's really only because I'm an old-timer that I'm able to explain that to
you off the top of my head, because since around 2004 and the rise of
journaled filesystems, fsck's job is so much easier (and quicker) that
nothing ever ends up in lost+found any more.

I'm sure there are atypical scenarios where it can happen. One is if you
suffer a thinko and run the _wrong fsck utiilty_, e.g., you ran fsck.ext2 on
an ext3 or ext4 filesystem, and it judged some or all existing files bad.
(Don't do that. ;-> ) Or, the hard drive hardware is failing and in the
process of failing creates invalid filesystem structures.

I do wonder what on Earth created that bizaarre outcome you mentioned -- but
it wasn't fsck.

[Scott DuBois]

Post by Rick Moen

Post by jim
$50.00 ---> $5.00

Oh, OK! Sorry, I didn't get the context. Sure, I like this; makes it more
like the gentleman's wager I intended. (Though, to be honest, it's most
fairly described as a cheap rhetorical device Iuse on rare occasions to
dramatise a point that I feared was getting lost in a messy Internet
discussion.)
Sure, if you want to do that. At a $5 level, I don't even mind if I'm
wrong, as it's a small price for learning something interesting.

Post by jim
I don't think fast power up-down will physically hurt the drive. I believe
I have a few drives lying around and if so will be happy to provide.

Bring the computer as well, and we have a deal. As I said, I think that
rapid a coldboot cycle stresses unduly the hard drive _and_ computer, and
I'm disinclined to throw away a PII just for a wager to make a point.

When and IF this goes down, I want to be invited to see the results; I'm
intrigued. :-)

[Rick Moen]

Post by bloonoise
Yeah as mentioned a good suggestion is to upgrade your power supply in
your box. I've also had problems with the cheap ones that come with the
machines. They can be problematic. But a simple upgrade solved all
problems I was having.

You betcha. Back when I was still using a workstation rather than laptops,
the smartest protective measure I ever took was to ditch the cheap-ass PSU
and substitute an Antec. (PC Power & Cooling was also great in the day,
arguably a couple of others.)

It should be noted that the cheap ones often severely failed to actually
come even close to their claimed wattage output capacity, leading to
the downstream devices overstressing the PSU, leading to catastrophic PSU
failure. Also of particular note, catastrophic PSU failure usually kills
the attached hard drives with a special murderous passion.

Post by bloonoise
Power supply's are pretty dam cheap on the net. Like 25 bucks for 450
watts.

But most of them are lying outrageously when they say they'll do 400 Watts.
Thus the problem, and the incentive to stick with the four or five actually
reputable brands that actually meet spec rather than fail and kill your hard
drives.

Post by bloonoise
SSDs are another upgrade well worth the investment.

Concur, though maxing out RAM is sometimes... actually, make that almost
always... a subtantially bigger bang for the buck.

People putting Linux on PeeCees are often all worried about whether the CPU
is powerful enough, because PeeCee advertising stresses CPU speed as if it
were the ne-plus-ultra determinant of performance. Which is hilarious,
because the typical PeeCee has bloated, out-of-proportion CPU performance to
meet MS-Windows's massive need for same -- and what typically makes a great
deal more difference is quantity of RAM. Linux can always put more RAM to
good use: As anyone who's spent time analysing 'ps' and 'free' output soon
figures out, if Linux isn't needing a chunk of RAM for anything else at the
moment, it gets roped into serving as disc cache. And disc cache, in turn,
makes mass storage really fly, as Linux does it very effectively.

Post by bloonoise
Now boot times are minimal, very.

People get all fixated on boot times, and I cannot fathom why. If your
Linux system is booting more than once every few months at most, then you're
doing something very wrong.

Obsessing over the speed of something that happens only once in a blue boom
seems kind of odd.

(I'm not saying _you_ are obsessing over it, and certainly having boot times
be shorter rather than longer is A Good Thing -- just not really high on
Maslow's Hierarchy of Needs, in my opinion.)

[bloonoise]

Yeah RAM is a worthy investment and a must
8 gigs is fine and enough, no more
4 OK I suppose
I dunno what the stats are for RAM versus sap on a SSD
but RAM is a must.

Post by Rick Moen
Concur, though maxing out RAM is sometimes... actually, make that almost
always... a subtantially bigger bang for the buck.

Yeah I only reboot after an update/grade
but its nice to have those speeds when you restart X after a crash.
And SSD's give the performance people expect from just about everything
else.
Even a new CPU or Mommy board or MeMory.
SSD's are great!

People get all fixated on boot times, and I cannot fathom why. If your
Linux system is booting more than once every few months at most, then you're
doing something very wrong.

[Scott DuBois]

Post by Rick Moen

Post by bloonoise
Yeah as mentioned a good suggestion is to upgrade your power supply in
your box. I've also had problems with the cheap ones that come with the
machines. They can be problematic. But a simple upgrade solved all
problems I was having.

You betcha. Back when I was still using a workstation rather than laptops,
the smartest protective measure I ever took was to ditch the cheap-ass PSU
and substitute an Antec. (PC Power & Cooling was also great in the day,
arguably a couple of others.)
It should be noted that the cheap ones often severely failed to actually
come even close to their claimed wattage output capacity, leading to
the downstream devices overstressing the PSU, leading to catastrophic PSU
failure. Also of particular note, catastrophic PSU failure usually kills
the attached hard drives with a special murderous passion.

Post by bloonoise
Power supply's are pretty dam cheap on the net. Like 25 bucks for 450
watts.

But most of them are lying outrageously when they say they'll do 400 Watts.
Thus the problem, and the incentive to stick with the four or five actually
reputable brands that actually meet spec rather than fail and kill your hard
drives.

Post by bloonoise
SSDs are another upgrade well worth the investment.

Concur, though maxing out RAM is sometimes... actually, make that almost
always... a subtantially bigger bang for the buck.
People putting Linux on PeeCees are often all worried about whether the CPU
is powerful enough, because PeeCee advertising stresses CPU speed as if it
were the ne-plus-ultra determinant of performance. Which is hilarious,
because the typical PeeCee has bloated, out-of-proportion CPU performance to
meet MS-Windows's massive need for same -- and what typically makes a great
deal more difference is quantity of RAM. Linux can always put more RAM to
good use: As anyone who's spent time analysing 'ps' and 'free' output soon
figures out, if Linux isn't needing a chunk of RAM for anything else at the
moment, it gets roped into serving as disc cache. And disc cache, in turn,
makes mass storage really fly, as Linux does it very effectively.

Post by bloonoise
Now boot times are minimal, very.

People get all fixated on boot times, and I cannot fathom why. If your
Linux system is booting more than once every few months at most, then you're
doing something very wrong.
Obsessing over the speed of something that happens only once in a blue boom
seems kind of odd.
(I'm not saying _you_ are obsessing over it, and certainly having boot times
be shorter rather than longer is A Good Thing -- just not really high on
Maslow's Hierarchy of Needs, in my opinion.)

To this I can say I typically focus on RAM bus speed and capacity when
shopping for mobo's. My current system runs a deprecated 1156 socket
core i5 with a capacity of 16G on the RAM to which I have only 8G of
Kingston HyperX 2133 bios set to 2K. At the time of build, the higher
capacity RAM was WAY out of budget but I still have upgrade potential.

http://www.anandtech.com/show/3558

http://us.hardware.info/productinfo/101802/kingston-hyperx--4gb-ddr3-2133-cl9-kit

My learning experience of RAM came in Y2K when my new desktop ran on a
66 bus and was too slow to perform many tasks my friends were doing
easily. More RAM and a faster RAM bus is always a priority, even before
processor specs. Other than reading articles, no one has really
explained the fine points so I've gathered on my own that RAM I/O and
MAX capacity is always a person's best bet.

Power is supplied from a "Black Widow" 550w. I've never been very
knowledgeable about PS's but the graphics card I have in the box stated
that it required a minimum of 500w so this one looked good from the time
I took to read reviews.

http://www.thermaltake.com/products-model.aspx?id=c_00001556

Primarily because of my decisions on what was "cool" factor in 2009, I
picked up the Antec 9001 case but now is a burden because of the LED
fans I can't turn off. It sits on my desk next to my bed and my wife
complains about the light it emits when trying to go to sleep. So, every
night gets shut down and every morning start up. I just need to get off
my ass and get a new case to resolve the problem.

(and the front fans make my right (mouse) hand cold which sucks)

http://www.altex.com/Antec-Nine-Hundred-9-Bay-Mid-Tower-Gaming-Chassis-900-P144693.aspx

Five years later this thing is still running pretty damn good
considering it gets shut down every night and runs ~16hrs/day.

[Rick Moen]

Post by bloonoise
SSD's are great!

They really are!

Post by bloonoise
Yeah I only reboot after an update/grade

Strictly speaking, that's useful only if the packages upgraded include your
running kernel. Otherwise, no need to rebot at all. This isn't MS-Windows.

[bloonoise]

Hmmmm don't new packages only run after a reboot?
maybe your right
certainly X or system files and such won't pull in until a reboot.
I use Arch principally which is constantly upgraded on a rolling release.
I dunno its been habit.
I certainly don't use Windblows(out my butt) much except virtually.
Also run fedora on two others which probably don't need a reboot much.
I dunno if I don't reboot I'll feel guilty and have to eventually......

Post by Rick Moen

Post by bloonoise
SSD's are great!

They really are!

Post by bloonoise
Yeah I only reboot after an update/grade

Strictly speaking, that's useful only if the packages upgraded include your
running kernel. Otherwise, no need to rebot at all. This isn't MS-Windows.
_______________________________________________
BALUG-Talk mailing list
http://lists.balug.org/listinfo.cgi/balug-talk-balug.org

[Rick Moen]

Post by bloonoise
Hmmmm don't new packages only run after a reboot?

Nope.

Post by bloonoise
Maybe you're right.

Hey, even a broken clock is right twice a day.

Post by bloonoise
certainly X or system files and such won't pull in until a reboot.

Nope.

If you have upgraded your running kernel, it is desirable to reboot as soon
as reasonably possible (unless using the KSplice utility, in which case, no
need to reboot at all). Otherwise, not needed or useful.

Post by bloonoise
I dunno its been habit.

I'm sure. Good news: You can stop doing that.

Post by bloonoise
Also run fedora on two others which probably don't need a reboot much.

Same story as upthread.

Post by bloonoise
I dunno if I don't reboot I'll feel guilty [...]

You can make pennance by donating to St. iGNUtius, I guess. ;->

[bloonoise]

actually this thread lead me to investigate.
and on Arch you need to logout/in for new packages

And you're right! only for a new kernel, or systemd, or glibc
do you need to reboot.
and those don't get updated a lot

Dunno about other Distro's yet but I think you're right.
Arch doesn't support partial upgrades as its constantly in motion
we get updates daily. And of course because of its nature it is a bit
less stable...but fun.

Post by Rick Moen

Post by bloonoise
Hmmmm don't new packages only run after a reboot?

Nope.

Post by bloonoise
Maybe you're right.

Hey, even a broken clock is right twice a day.

Post by bloonoise
certainly X or system files and such won't pull in until a reboot.

Nope.
If you have upgraded your running kernel, it is desirable to reboot as soon
as reasonably possible (unless using the KSplice utility, in which case, no
need to reboot at all). Otherwise, not needed or useful.