Slashdot is powered by your submissions, so send in your scoop

 



Forgot your password?
typodupeerror

Comment: Re:Mandatory xkcd (Score 1) 229

by rev0lt (#49557043) Attached to: GNU Hurd 0.6 Released

Still an order of magnitude more complex than systemd.

Actually, its not. In FreeBSD, the /etc/rc.conf has a somewhat similar setup to enable/disable services ,as other Linux distros also have. Those you shown are global parameters, not startup scripts.

Then what are we discussing? Use your old style sysvinit.

Usually you can't just disable SystemD, and the tendency (IE it WILL happen), sysvinit will be removed. Try CentOS or Debian w/o SystemD and let me know how well it goes.

I don't know again what you mean. The sysvinit init scripts are still working.

They aren't, at least not fully. This requires an extra step in SystemD, to enable legacy processing of rc.d scripts.

The journald is used internally by systemd, and services can opt-in to use it.

Its not really op-in, is it? if SystemD generates diagnostic messages to journald instead of syslog or whatever you're using, you need to use it at least in some point. And when shit happens, systemD instructs you to use some crazy command to parse the binary logs, not a "hey this program is logging elsewhere". So, from a systemd user perspective, there is no other logging facility - error messages instruct you to parse journald, not to check the application log in /var/whatever.

At least systemd captures sysout and syserr into the journal, so it's not lost

Precisely. Its not opt-in. And when you manually startup scripts, it captures the useful information the program is outputing *at that instance*.

Another huge improvement over sysvinit. How many times I had to start a daemon manually so I can see the errors from syserr.

Most daemons will log the relevant messages. And its actually trivial to enable logging of those on your startup script, if you want it (and some programs, like crontab, already have them). What we have now, are deamons silently failing and requiring an extra command to understand why.

But again, the MySQL logs are still there in /var/log ? It's not systemd fault if Oracle is not updating MySQL to use journald.

No, they aren't. In many Linux distros MySQL is configured in /var/lib. And its not Oracle's fault that someone implemented a startup script that is dead useless with non-supported applications. At least sysvinit works with everything.

Comment: Re: Dell, HP, Panasonic (Score 1) 417

by rev0lt (#49552823) Attached to: We'll Be the Last PC Company Standing, Acer CEO Says
So does Belkin, D-Link, TP-LINK, Genius and a bunch of other "virtual" manufacturers. Case design is different, sometimes the PCI is different, but the circuit is the same. And more often than not, even the firmware is the same. The case of ASUS isn't always this (often they provide changes to the base circuit or a customized firmware), but generically is a whilelabel brand. They sell re-branded stuff assembled by others.

Comment: Re:Mandatory xkcd (Score 1) 229

by rev0lt (#49552783) Attached to: GNU Hurd 0.6 Released

Then you want systemd. As I mentioned before, the sysvinit start scripts are huge, complicated and error prone (the loop in Apache init script). In systemd all you have is a 17 lines file for Nginx: https://projects.archlinux.org... [archlinux.org]

Your argument is a bit flawed. You can actually bootstrap apache with a single command (apachectl start) you can include in /etc/rc.local or similar startup script. For nginx, it gets a bit more complex, as no deamon control program exists (while you can actually do most stuff by using the -s modifier). So, comparing apples to apples, the apache22 init script in FreeBSD has around 190 lines, and nginx's ~140, including comments and dependencies. The scripts are quite easy to understand and show how to put everything in place. Ever tried to run multiple copies of the same program simultaneously with completely different parameters? With a shellscript, is trivial to do so.

systemd adds a lot of extras, and does not take anything away. It adds, a) info on whether the service was started, b) automatic restart of crashed services, c) journald, d) limit resource usage per service, e) simple and powerful service configuration files, f) inter service dependencies management (start Thin only if Nginx is started, or start MySQL after Apache) g) a lot of tools that were missing like systemd-ask-password.

...None of them are useful to me. a) Info if the service is started or not is standard in every rc.d/init.d system I know, via the service command or other methods. b) the LAST thing I want is automatic restart of crashed services. I want to know it crashed, and investigate why (automatic restart easies the process of remote code execution in network daemons). c) journalD adds nothing useful. If I wanted my logs in a database, pretty much half of the loggers available have that option. d) this already exists in (every?) unix system. Its called process accounting and limits. Dependency managent is a solved problem - either via explicit ordering as in traditional linux systems, or explicit dependency enumeration in FreeBSD/NetBSD. If you think this is a cool feature, you've been living under a rock for the past decade or so. g) systemd-ask-password is a prime example of a cool feature that has *nothing* to do with the init process, so it should not be part of it.

And systemd does not take anything away from you. You can still use all your old school tools and your old style logging, and even the old sysvinit scripts.

Theoretically. In practice, it overrides the way sysvinit works, and you need to explicitly re-enable it.

"but if not, why the f*ck is it taking control over MySQL initialization? " What do you mean by that? It starts the service and watches the service that it is really started, and it gives you info that it is started or it didn't started or it crashed. For example, on my laptop:

Try it with MySQL. Your output actually has useful info, in MySQL's case, it doesn't, as systemd isn't capturing the program's log. And what if I had multiple mysql copies running?
In the end, systemd adds nothing to the existing process. Maybe stuff boots faster on a laptop, but I don't really care about startup times. I care about reliability and being able to fix stuff when it breaks. Again, systemd stands in the way, its not a help. If you feel intimidated by a 9Kb shellscript file, It may be suitable for you, but it's not for me. I prefer to dig around a small shellscript file than to have to read manpages and google stuff dispersed on the internet.

Comment: Re:Dell, HP, Panasonic (Score 1) 417

by rev0lt (#49552677) Attached to: We'll Be the Last PC Company Standing, Acer CEO Says

You can take your MSDOS and boot it on your new i7 computer.

You're mixing ISA (Instruction Set Architecture) with ISA (Industrial Standard Architecture). Two completely different things.

you could easily make a LPC to ISA adaptor in a FPGA

Good for you. I'd be a bit skeptical about that (at least with full compatibility, including DMA & IRQ support), but never tried it.

But I guess you are trying to insert some non-existent expert knowledge or something?

I could say exactly the same about you. The difference is, I don't automatically assume everyone is wrong, nor that they are assholes. Different perspectives, I guess.

Comment: Re: Dell, HP, Panasonic (Score 2) 417

by rev0lt (#49542949) Attached to: We'll Be the Last PC Company Standing, Acer CEO Says
Asus doesn't "build" computers :) Foxconn does. And also does Apple and Intel gear (AFAIK). And on the laptop market, at least some years ago, Quanta (Taiwan) had allegedly assembled 30% of all the sold laptops. You can usually spot this by checking the MAC address on the ethernet card (see http://www.coffer.com/mac_find... ).

Comment: Re: Dell, HP, Panasonic (Score 1) 417

by rev0lt (#49542927) Attached to: We'll Be the Last PC Company Standing, Acer CEO Says
Its still the same. Most boards use a variant of the Intel reference design, with usually a variable quality choice in components. Cheaper boards sometimes even have the same silklayer as Intel boards. Intel used to (at least a decade ago) to provide full schematics for their chipset combos.

Comment: Re:Dell, HP, Panasonic (Score 1) 417

by rev0lt (#49542915) Attached to: We'll Be the Last PC Company Standing, Acer CEO Says

Also referred to in the past as the ISA (Industrial Standard Architecture)

ISA is actually a set of standards. Many/most XT ISA boards won't work in AT systems and vice-versa. Your parallell port may work, but a SCSI card may not.

PCs still are compatible with the old standard

Hum, no. ISA has been dead at least since the Pentium IV days, and even then it would require extra circuitry to use it. Maybe some modern industrial PCs still have it, but I'd guess its mostly emulated. DMA and IRQs are handled differently, bus timings and AFAIK voltages are different.

Comment: Re:Mandatory xkcd (Score 1) 229

by rev0lt (#49542757) Attached to: GNU Hurd 0.6 Released
My problem is simple. Why am I using a tool that adds nothing of what I need, and - on the other hand - abstracts a ton of simple stuff you could easily see on a shellscript on an obvious location? Would it be that hard to just say "hey it went tits up. I have no info why, check the MySQL log"? No, it will tell you to go trough all those hoops you suggested (after all, it may just be an initialization problem, right?) before you realize you need to look elsewhere.
With netstat you don't even need to know the service name (I actually never use servicenames). There is a table for that in most (all?) unix systems, in /etc/services. Your ubuntu server won't have a process called nginx.conf. It will have a process called nginx, that may or may be not initialized with a configuration file. I'm quite familiarized with the standard unix toolset (15 years of several BSD flavours will do that to you), but no standard tool will help you when some opaque software is capturing the initialization process. Do you know which flags are passed to eg. your nginx setup? A quick look at the rc.d script will enighten you. Is there a bug on the script (it happens more often than you seem to think)? Easy to fix. I know that SystemD still uses shellscripts on a convoluted structure, but how is this better? Hint - it's not.
What I want is quite simple. To be able to configure and manage stuff with the least hassle. SystemD is a stone in my shoe, not a solution. And its not because I'm BSD-centric (I am), its because it hides complexity AND useful info. I'm fine if MySQL isn't logging into SystemD, I really don't care - but if not, why the f*ck is it taking control over MySQL initialization?
I've worked with SMF in Solaris. Coming from a BSD background, its a bit of a steep learning curve, because it works quite differently from the init.d/rc.d approach. Some operations are overly complex, some ways of doing stuff are convoluted, but I could see the added value in the tool. I see none in SystemD.

Comment: Re:Mandatory xkcd (Score 1) 229

by rev0lt (#49538437) Attached to: GNU Hurd 0.6 Released
So, your suggestions are straight out of SystemD manual (and actually printed on console when a service fails) and you call my issues with SystemD too unspecific?
I actually mentioned MySQL, whose errors are not actually related to the kernel (so dmesg is completely useless), and if a service is *not* running, so it is ps or netstat and any of your suggestions. Try configuring a server, have a typo on some obscure parameter, and see the difference between the MySQL log and what SystemD actually spews out. Its great if it works for you, but I actually require useful info beyond the "oops it failed" to do my job.

Comment: Re:Mandatory xkcd (Score 1) 229

by rev0lt (#49497597) Attached to: GNU Hurd 0.6 Released

"# wait until really stopped"

Or just use apachectl stop or whatever. With systemD, I get ZERO useful info when eg. MySQL doesn't start, but I do receive indication to run some absurd command to tell me the details, and the details are mostly "bad luck dude, it didnt start" in fancy language. Because apparently circumvents the traditional /var/log/messages logging system, you get ZERO details for system-related failures. You'll end up using the MySQL error log, which is also in a funny location in many linux installations (why the fuck MySQL is in some lib folder?).
I'm forced to use Linux in some boxes and - truth be told - I'm no Linux fan - but SystemD makes sure the experiment is excruciating. Want to quickly add some routes on bootup? Don't forget to enable the rc,* startup script, because "fuck you thats why - lets ditch decades-old compatibility by requiring you to run extra commands". Half the time I spend configuring setups is messing around with SystemD and why the fuck it truncated the application-generated error messages. Everytime I have to mess with it I feel the urge to punch someone.

"Life sucks, but it's better than the alternative." -- Peter da Silva

Working...