That's all fine and well, but inittab doesn't have enough smarts to start and stop myriad interdependent processes in order. It also doesn't offer you any way to specify how to handle the respawing of various processes with different requirements, etc., etc.
It would be fine with me if good old init got upgraded, and would do the job well, but it hasn't been, instead, at least twice, groups that wanted to improve it, started over from scratch, and they have something that works.
With a few thousand servers, basic, system level services (that have run fine for months) will be crashing unexpectedly (somewhere on the cluster) every day. And that even with absolutely nothing wrong with the systems. And when you "simply [re]start" it, it will continue to run for several more months without any trouble at all. At least, on heavily-utilized clusters.
And be slow as shit and not work properly, which was the state of video in the browser under linux since... forever, pretty much. I was about to say Flash fixed it, but it has all those problems and worse, sadly.
Video in the browser on Linux worked great with MPlayerplug-in... Vastly better than using the proprietary browser plug-ins for video on Windows.
It was great, but Flash destroyed it, and took both platforms a decade backwards... Flash finally got the "hardware acceleration" that other browser plug-ins had from the start, but still sucks.
you can simply run/etc/rc every minute via cron and it'll sync what's running with what's supposed to be assuming things have been/sbin/service stopped.
A "crash" does not involve anyone running "service X stop" so you're providing a solution for a problem nobody wants nor asked about.
(And if they haven't been cleanly stopped, you need a specialized tool that understands how to *TEST* the service rather than rely on subsys.)
And that specialized tool is... wait for it...
THE SERVICE'S OWN INIT SCRIPT!!!
That's right, you can iterate through service "$X" status on everything, and do a restart on anything that has terminated, but that's just a hack, and something that can be done infinitely superior within the software handling the service startup... namely, upstart or systemd.
I hate Lennart Poettering and PulseAudio as much as anybody, but SysVinit is broken, and systemd is a fix. Angry zealots repeatedly denying that there's a problem, is probably why it's taken so very long before a fix finally arrived.
I'm leery of systems that automatically restart services when they crash, especially if the service just crashes again at startup, and you get into an infinite loop that eventually runs you out of disk space with *.core files.
It takes only a trivial amount of brains to prevent such loops, even Windows NT svc does this fine.
Then you have time to investigate why things are failing on one node, and can implement a proper fix.
You aren't going to investigate why crond died one time, after running for several months straight, if it can just be restarted and run for several more months without problems. With enough servers, you'd need a dedicated employee spending all day doing nothing else, and almost never finding anything that can be fixed. If a service keeps crashing, THEN you can investigate.
It's almost always something transient and trivial like a DNS server just didn't happen to respond in-time. Investigating all such occurrences is stupid, offers no return, and the situation is made 10X worse when you get a paged at 3AM every single night, because there will always be some random box in the cluster with some transient failure of one service.
Auto-restarting crashing daemons is not a feature. It's a band-aid over top of poor system administration.
Only someone who has never administered 1,000+ servers would say that. Sooner or later, the most reliable services are going to crash. With hundreds of servers, you'll see it several times every day.
DJB was no fool when he included the feature in DaemonTools, and he didn't do it because his software was crap, or because he doesn't know how to administer a system. The same goes for the people behind every major Linux distro, who have decided to adopt systemd. They aren't doing it for shits and giggles. And desktops make up such a tiny percentage of RedHat's sales that it's ridiculous to claim they'd make any changes that aren't for the benefit of their big server customers.
Telling other people THEY shouldn't have a feature that YOU don't see the need for, is the crux of this whole SystemD holy-war and flame-fest.
All of your substantive complaints seem to be a direct result of ignoring the principle "don't fix what isn't broken".
SysVInit is broken. It doesn't restart services that crash. In an environment with many hundreds of servers, the menial task of restarting services becomes a full-time job.
- Useless on a server - where you only reboot 4 times a year or so and never have to hot-plug anything or change wireless networks.
Bull. Lots of servers currently run daemontools or similar, or else they use some other hack, because the SysVinit doesn't have any way to restart services (like crond) the one time they exit after running fine for months...
Alternatively, somebody has to take the time to set-up exhaustive monitoring, including ALL the trivial services running on the servers, and some dummy has to watch it around the clock, and manually perform this extremely simple and menial task. Or else maybe you're the dummy who gets paged at 3AM to do a trivial service restart, due to some simple and transitory event.
I would have been just as happy with upstart or anything else, but it was a dammed nuisance lacking that 30 year-old feature, and downright embarrassing that Linux still lacked it, while it's been working well in the base of Windows since the first version of NT.
The only time you have to restart Linux is if you change Kernel or Kernel header files. Otherwise just upgrade and walk away without having to reboot.
Does apt-get (or yum) automatically restart every service and program that uses a library that it has updated?
No? Then you can't just 'walk away' if you care about security in the slightest. Your running Apache will continue using the buggy old OpenSSL version until you restart the service. You *could* take your system down to single-user "emergency" mode, then back up... That's technically not a reboot, but close enough that most people would call it that.
we're worried about dying from Global Warming . . . getting hit by an asteroid . . . an Ebola epidemic . . . but nobody seems concerned that maybe the Earth could bust apart at its seems.
You're kidding, right?
Just after people's terror of word-ending asteroids wore off, the media was pushing the Yellowstone Supervolcano (very hard) as the thing we should all be pissing our pants about. And they really never gave-up on it, either:
Your anecdotal experience is directly contradicted by lots of other anecdotal experience out there.
Certainly no manufacturers claim their tapes can hold data reliably for decades, particularly since the repeated-contact and non-solid-state nature of reel-fed tape makes multi-use reliability basically impossible.
Meanwhile, Sony certainly did advertise and guarantee the reliability of their WORM MO disc technology for multi-decade archival purposes.
MO disks require (IIRC) a bit to be raised to a very high temperature to alter, while bluray just requires the organic dye to degrade (as they all do).
There are at least 3 distinct types of Blu-ray discs: Commercially pressed, -R, and -RW (well, they call them -RE, but... meh).
Only one of the three types uses an organic dye that degrades. Instead BD-RW has much in common with MO discs, and was reportedly the first format Sony developed, thanks to their existing MO technology.
You'd have to be ignorant or foolish to rely on dye-based mediums like bluray for anything archival.
You are sadly showing your ignorance of disc technology. I've handled enough of both in my time to make a far better judgment than an armchair expert.
Enterprises have been doing this with tape for 30 years.
Tape has always had a limited life-span and is too easily damaged to completely trust with high-value archival data. Instead, archival on tape usually means "we're not quite confident enough to just delete this crap".
Meanwhile, Sony's enterprise-grade write-once (WORM) magneto-optical (MO) discs have been around for decades, are physically tougher, and impervious to magnetic fields, sold with 100-year warranties that even cover data-loss recovery costs.
BD-RW can certainly be seen as Sony's MO technology being brought down dramatically in price due to economies of scale, and intentionally to allow them to compete in the consumer space.
Norway at first place has 6.1% penetration, followed by five other European countries and Japan, while USA is down at 8th place, with an order of magnitude(!) less electric car penetration than Norway.
Those are bullshit numbers. That's percentage OF CAR SALES that are EVs. Areas with fewer cars to begin with are disproportionally represented.
In addition, it's BS to compare a huge country, with a bunch of small ones. There will be both extremes in the statistical distribution. If we carved-up the US into similarly-sized pseudo-countries, there would be areas with extremely high EV penetration, and they wouldn't get penalized for the parts of the US with extremely low penetration.
And just above that table:
"Plug-in hybrid sales in 2012 were led by the United States with a 70% share of global sales, followed by Japan with a 12%, and the Netherlands with 8%."
Per-capita, Japan may be ahead of the US, but Europe sure as hell isn't.
Slashdot Mirror
That's all fine and well, but inittab doesn't have enough smarts to start and stop myriad interdependent processes in order. It also doesn't offer you any way to specify how to handle the respawing of various processes with different requirements, etc., etc.
It would be fine with me if good old init got upgraded, and would do the job well, but it hasn't been, instead, at least twice, groups that wanted to improve it, started over from scratch, and they have something that works.
With a few thousand servers, basic, system level services (that have run fine for months) will be crashing unexpectedly (somewhere on the cluster) every day. And that even with absolutely nothing wrong with the systems. And when you "simply [re]start" it, it will continue to run for several more months without any trouble at all. At least, on heavily-utilized clusters.
Video in the browser on Linux worked great with MPlayerplug-in... Vastly better than using the proprietary browser plug-ins for video on Windows.
It was great, but Flash destroyed it, and took both platforms a decade backwards... Flash finally got the "hardware acceleration" that other browser plug-ins had from the start, but still sucks.
A "crash" does not involve anyone running "service X stop" so you're providing a solution for a problem nobody wants nor asked about.
And that specialized tool is... wait for it...
THE SERVICE'S OWN INIT SCRIPT!!!
That's right, you can iterate through service "$X" status on everything, and do a restart on anything that has terminated, but that's just a hack, and something that can be done infinitely superior within the software handling the service startup... namely, upstart or systemd.
I hate Lennart Poettering and PulseAudio as much as anybody, but SysVinit is broken, and systemd is a fix. Angry zealots repeatedly denying that there's a problem, is probably why it's taken so very long before a fix finally arrived.
grep -ve '^$' -e '^#' ldap.conf
Did you say that when web browsers started rendering images, and included javascript support?
It takes only a trivial amount of brains to prevent such loops, even Windows NT svc does this fine.
You aren't going to investigate why crond died one time, after running for several months straight, if it can just be restarted and run for several more months without problems. With enough servers, you'd need a dedicated employee spending all day doing nothing else, and almost never finding anything that can be fixed. If a service keeps crashing, THEN you can investigate.
It's almost always something transient and trivial like a DNS server just didn't happen to respond in-time. Investigating all such occurrences is stupid, offers no return, and the situation is made 10X worse when you get a paged at 3AM every single night, because there will always be some random box in the cluster with some transient failure of one service.
Only someone who has never administered 1,000+ servers would say that. Sooner or later, the most reliable services are going to crash. With hundreds of servers, you'll see it several times every day.
DJB was no fool when he included the feature in DaemonTools, and he didn't do it because his software was crap, or because he doesn't know how to administer a system. The same goes for the people behind every major Linux distro, who have decided to adopt systemd. They aren't doing it for shits and giggles. And desktops make up such a tiny percentage of RedHat's sales that it's ridiculous to claim they'd make any changes that aren't for the benefit of their big server customers.
Telling other people THEY shouldn't have a feature that YOU don't see the need for, is the crux of this whole SystemD holy-war and flame-fest.
SysVInit is broken. It doesn't restart services that crash. In an environment with many hundreds of servers, the menial task of restarting services becomes a full-time job.
Bull. Lots of servers currently run daemontools or similar, or else they use some other hack, because the SysVinit doesn't have any way to restart services (like crond) the one time they exit after running fine for months...
Alternatively, somebody has to take the time to set-up exhaustive monitoring, including ALL the trivial services running on the servers, and some dummy has to watch it around the clock, and manually perform this extremely simple and menial task. Or else maybe you're the dummy who gets paged at 3AM to do a trivial service restart, due to some simple and transitory event.
I would have been just as happy with upstart or anything else, but it was a dammed nuisance lacking that 30 year-old feature, and downright embarrassing that Linux still lacked it, while it's been working well in the base of Windows since the first version of NT.
Ah, good old Freudian-bullets... That's LibreOffice's killer feature over MS Word, you know?
Does apt-get (or yum) automatically restart every service and program that uses a library that it has updated?
No? Then you can't just 'walk away' if you care about security in the slightest. Your running Apache will continue using the buggy old OpenSSL version until you restart the service. You *could* take your system down to single-user "emergency" mode, then back up... That's technically not a reboot, but close enough that most people would call it that.
I'm sure there are people on Reddit (or Digg, or StumbleUpon) saying the same thing about stories posted to /.
A gigantic crater with lava flowing through it, will look an awful lot like something "split the Earth apart" to most people.
You're kidding, right?
Just after people's terror of word-ending asteroids wore off, the media was pushing the Yellowstone Supervolcano (very hard) as the thing we should all be pissing our pants about. And they really never gave-up on it, either:
http://www.inquisitr.com/10848...
http://www.bbc.com/news/scienc...
http://www.cnn.com/2012/08/30/...
http://news.nationalgeographic...
http://www.imdb.com/title/tt04...
And when headlines read "Southern California" most people think of Los Angeles...
Why should headlines cater to stupid people?
Your anecdotal experience is directly contradicted by lots of other anecdotal experience out there.
Certainly no manufacturers claim their tapes can hold data reliably for decades, particularly since the repeated-contact and non-solid-state nature of reel-fed tape makes multi-use reliability basically impossible.
Meanwhile, Sony certainly did advertise and guarantee the reliability of their WORM MO disc technology for multi-decade archival purposes.
I never understood what motivates pedants to waste everyone's time when nobody else cares.
The relevant phrase is an old journalism chestnut:
"Don't bury the lead!"
http://laurabrowncommunication...
There are at least 3 distinct types of Blu-ray discs: Commercially pressed, -R, and -RW (well, they call them -RE, but... meh).
Only one of the three types uses an organic dye that degrades. Instead BD-RW has much in common with MO discs, and was reportedly the first format Sony developed, thanks to their existing MO technology.
You are sadly showing your ignorance of disc technology. I've handled enough of both in my time to make a far better judgment than an armchair expert.
LTO didn't even exist 30 years ago, so your experience says little to nothing about archival quality.
Tape has always had a limited life-span and is too easily damaged to completely trust with high-value archival data. Instead, archival on tape usually means "we're not quite confident enough to just delete this crap".
Meanwhile, Sony's enterprise-grade write-once (WORM) magneto-optical (MO) discs have been around for decades, are physically tougher, and impervious to magnetic fields, sold with 100-year warranties that even cover data-loss recovery costs.
BD-RW can certainly be seen as Sony's MO technology being brought down dramatically in price due to economies of scale, and intentionally to allow them to compete in the consumer space.
Europe and the EU has many more people, yet VASTLY lower EV and plug-in hybrid sales.
That's complete crap, no matter how many times you repeat it. The figures quoted are for domestic sales, and do not count exports at all.
How is this different from the last time the topic was on the front-page of /.?
http://hardware.slashdot.org/s...
Those are bullshit numbers. That's percentage OF CAR SALES that are EVs. Areas with fewer cars to begin with are disproportionally represented.
In addition, it's BS to compare a huge country, with a bunch of small ones. There will be both extremes in the statistical distribution. If we carved-up the US into similarly-sized pseudo-countries, there would be areas with extremely high EV penetration, and they wouldn't get penalized for the parts of the US with extremely low penetration.
And just above that table:
"Plug-in hybrid sales in 2012 were led by the United States with a 70% share of global sales, followed by Japan with a 12%, and the Netherlands with 8%."
Per-capita, Japan may be ahead of the US, but Europe sure as hell isn't.
Both Europe and the EU proper have considerably more people than the US, so they're WAY behind per-capita, as well.
If you don't like my source, you're free to provide your own to backup your ridiculous claims, but I don't expect you will...