This systemd thing is really out of hand.

Polygeekery

This is less true with new OS versions.

Fair enough, but I am talking about Win7, Server 2008R2 and Server 2012R2.

Vista and XP just completely shit themselves at around 30 days, so I obviously was not talking about them. ;)

FrostCat

@Intercourse said:

Vista and XP just completely shit themselves at around 30 days

I have a machine in the office that i know goes longer than that. To be fair, it mostly sits unused. It DOES seem to get slower over time, but since it's only got a single-core Pentium 4 and a gig of ram, that might be the problem there, and not Windows, if you know what I mean.

Polygeekery

One of our clients uses a tons of legacy software and we have migrated the old P4 machines it was running on to VMs and upped the RAM on them. Generally, they are sitting idle. Users RDP in to the VMs, do their work, and log back out. After 30 days of uptime they are so slow as to be unusable.

Getting users to reboot a VM through RDP was difficult, so each machine has a .bat file on their desktop which contains:

shutdown -t 0 -r -f

So, it does not completely shit itself. It does become unusable though, even with most of them running 2 cores and 3GB of RAM.

dkf

@FrostCat said:

I didn't say locking, I said serialization. You have one thread responsible for all writes; other threads hand it text and say 'please add this to the log file'. The one thread does that, in order.

Whereas if there was an equivalent to the POSIX O_APPEND flag, that write serialization could be done in the kernel, where it is simpler to get right (especially with networked filesystems, but that's a whole 'nother story). On all true POSIX systems, as long as you write() entire records at a time, your logfile won't become corrupted (except for hardware failures and kernel bugs) if you're using the O_APPEND option, since it requires a synchronous seek-to-end as part of the write() call. It makes logging to a single file from multiple processes or multiple threads trivial, and I wish Microsoft would incorporate something equivalent because it really is that simple for client code.[spoiler]Note that MS does do other bits better, such as thread support. This just happens to be a weakness.[/spoiler]

Bort

Yeah, Windows never crashes so long as you don't do something idiotic like install any third-party software or something unforgivably retarded like run it on hardware.

ben_lubar

The latter also applies to self-driving cars.

dkf

@Bort said:

Yeah, Windows never crashes so long as you don't do something idiotic like install any third-party software or something unforgivably retarded like run it on hardware.

Yep, and from this perspective, packages like Microsoft Office are third-party software.

Windows doesn't crash if you never power the machine up. It must be the fault of that pesky electricity!

ender

@blakeyrat said:

All OSes crash when paired with unreliable broken hardware. That's the only reason Windows NT crashes, and has been for... a solid decade now.

That Server 2008R2 Crash I mentioned? It's an internal webserver, running a single low-importance webapp, and it doesn't even have antivirus installed. ATTEMPTED_WRITE_TO_READONLY_MEMORY in netbt.sys. There's zero non-Microsoft drivers installed (it's running on Hyper-V after all).

BTW, the most stable Windows I've seen so far is the Windows 98 VM we have at work that's running the door access control software (which doesn't run on anything NT-based, and is too costly to replace). It had around 3 years of uptime until October last year (when we had a planned extended power outage), and is now again at over a year of uptime.

@Intercourse said:

How the fuck do you want me to prove it? Windows machines become unstable after 60 days. They slow down appreciably.

Huh? I've seen plenty of machines with way over 60 days of uptime, without any such symptoms (hell, I'm sometimes lazy at updating my work machine, and it gets a few update cycles behind without being powered down - that was true with XP, and then 7, and 8 and now 8.1). I distinctly remember the NT4 server that had 245 days of uptime when I started working here (then it bluescreened when I tried opening Notepad), and I've seen plenty of clients with SBS servers that were never updated or rebooted before we started working on them.

FrostCat

@ender said:

That Server 2008R2 Crash I mentioned? It's an internal webserver, running a single low-importance webapp, and it doesn't even have antivirus installed. ATTEMPTED_WRITE_TO_READONLY_MEMORY in netbt.sys. There's zero non-Microsoft drivers installed (it's running on Hyper-V after all).

RAM failure? Have you tried running something like memtest86 on the host OS?

ender

@FrostCat said:

RAM failure? Have you tried running something like memtest86 on the host OS?

ECC RAM, nothing about (un)correctable errors in Event Viewer on host.

flabdablet

Hadn't seen zgrep before - thank you.

flabdablet

@Intercourse said:

That little anecdote illustrates how Windows chokes with long uptimes and also just how damned incompetent some sysadmins are.

I used to have a scheduled task on both the school servers (Windows Server 2003) to reboot them every night at 2:30am, which I added because without it they used to get really, really slow or even lock up completely after four or five days. No idea why.

That task is gone now because I've since p2v'd both those boxes and the host takes them down every 24 hours to snapshot itself and back everything up, so they're getting regular reboots without needing to do it themselves.

flabdablet

@Intercourse said:

>boomzilla:
I agree. It's like my unhealthy obsession with a rock when it gets into my shoe.

I would have went with hemorrhoid.

I don't think I've ever had a haemorrhoid in my shoe. That sounds messy.

Scarlet_Manuka

@flabdablet said:

I don't think I've ever had a haemorrhoid in my shoe. That sounds messy.

I was going for that as well when I read it this morning, but after composing my reply nothing happened when I pressed the button to post it. Glad to see the TDWTF community was there to take care of it.

FrostCat

@flabdablet said:

I used to have a scheduled task on both the school servers (Windows Server 2003) to reboot them every night at 2:30am, which I added because without it they used to get really, really slow or even lock up completely after four or five days. No idea why.

Probably because students abuse the lab computers. When I was in college they were constantly being reimaged.

ijij

@blakeyrat said:

Teddy Roosevelt in a Chevy Nova V-8

Want.

flabdablet

@FrostCat said:

Probably because students abuse the lab computers

Nope. Primary school (rough US equivalent: K-6). And one of the crashy servers was the admin server, connected only to four admin workstations and a Cisco internet router.

The only software running on both was Windows Server 2003.

VaelynPhi

@flabdablet said:

Meh again. ls -rt /var/log/syslog* | xargs zcat | grep blah | less gets the job done, and even if I'm working on a box where I don't have that defined as gl() it takes less time to type than to get Event Viewer up and wait for it to sort itself out.

Didn't work for me; complained that the first two logs weren't gzipped.

@PJH said:

zgrep blah /var/log/syslog* | less

This also did not work for my first two logs.

@PJH said:

Either way, don't grep your cats.

Mine enjoys being grepped, especially with the pick that removes stray hair.

tarunik

@VaelynPhi said:

Didn't work for me; complained that the first two logs weren't gzipped.

This is normal -- rsyslog doesn't gzip any given log file until it's been rotated once or twice. If it happened anytime recently, just grepping /var/log/syslog does the trick, anyway.

flabdablet

@VaelynPhi said:

> flabdablet:
Meh again. ls -rt /var/log/syslog* | xargs zcat | grep blah | less gets the job done, and even if I'm working on a box where I don't have that defined as gl() it takes less time to type than to get Event Viewer up and wait for it to sort itself out.

Didn't work for me; complained that the first two logs weren't gzipped.

Sorry, brain fart. Should have been xargs zcat -f

FrostCat

@flabdablet said:

Nope. Primary school (rough US equivalent: K-6).

Hmm. Well, I did say my experience of seeing it was in college.

flabdablet

Also can't find any option to make zgrep fall back to grep for non-compressed input, so I guess I'll stick with my existing zcat -f | grep recipe.

PJH

@flabdablet said:

Also can't find any option to make zgrep fall back to grep for non-compressed input

Because it does that by default anyway - at least every version I've ever used has...

[root@sofa ~]# zgrep CROND /var/log/messages | tail
Nov 25 17:09:01 sofa CROND[5635]: (root) CMD ([ -d /var/lib/php ] && find /var/lib/php/ -type f -mmin +$(/usr/lib/php/maxlifetime) -print0 | xargs -r -0 rm)
Nov 25 17:39:01 sofa CROND[7074]: (root) CMD ([ -d /var/lib/php ] && find /var/lib/php/ -type f -mmin +$(/usr/lib/php/maxlifetime) -print0 | xargs -r -0 rm)
Nov 25 18:01:01 sofa CROND[8144]: (root) CMD (nice -n 19 run-parts --report /etc/cron.hourly)
Nov 25 18:09:01 sofa CROND[8551]: (root) CMD ([ -d /var/lib/php ] && find /var/lib/php/ -type f -mmin +$(/usr/lib/php/maxlifetime) -print0 | xargs -r -0 rm)
Nov 25 18:39:01 sofa CROND[9987]: (root) CMD ([ -d /var/lib/php ] && find /var/lib/php/ -type f -mmin +$(/usr/lib/php/maxlifetime) -print0 | xargs -r -0 rm)
Nov 25 19:01:01 sofa CROND[11105]: (root) CMD (nice -n 19 run-parts --report /etc/cron.hourly)
Nov 25 19:09:01 sofa CROND[11508]: (root) CMD ([ -d /var/lib/php ] && find /var/lib/php/ -type f -mmin +$(/usr/lib/php/maxlifetime) -print0 | xargs -r -0 rm)
Nov 25 19:39:01 sofa CROND[12969]: (root) CMD ([ -d /var/lib/php ] && find /var/lib/php/ -type f -mmin +$(/usr/lib/php/maxlifetime) -print0 | xargs -r -0 rm)
Nov 25 20:01:01 sofa CROND[14035]: (root) CMD (nice -n 19 run-parts --report /etc/cron.hourly)
Nov 25 20:09:01 sofa CROND[14440]: (root) CMD ([ -d /var/lib/php ] && find /var/lib/php/ -type f -mmin +$(/usr/lib/php/maxlifetime) -print0 | xargs -r -0 rm)
[root@sofa ~]# file /var/log/messages
/var/log/messages: ASCII text
[root@sofa ~]#

flabdablet

stephen@kitchen:~$ echo foo | zgrep .
foo
stephen@kitchen:~$

schweet!

VaelynPhi

@tarunik said:

This is normal -- rsyslog doesn't gzip any given log file until it's been rotated once or twice. If it happened anytime recently, just grepping /var/log/syslog does the trick, anyway.

I know; this is the log wrangling I was originally pointing out had to be done.

regex="local[a-z]*"
log="/var/log/syslog"
zgrep -e $regex $log.* | sed "s|^[^:]*:||g" | sort -r

Unfortunately, the sed is necessary because zgrep lies about taking all of grep's options (the option for suppressing the file being printed does not work, though perhaps it has a different switch and I just need to check the source).

In other words, there is no straightforward way to search the logs; it requires some scripting to be able to do simply.

@flabdablet said:

Sorry, brain fart. Should have been xargs zcat -f

Voila. This did work. Thanks. And it's rather convenient too (keeping the search term on the right). I looked, though, and all the z* utilities are scripts, so pretty much not terribly better than something homegrown, but at least standard.

ben_lubar

riking

...

http://i.imgur.com/XDvufBu.png

ben_lubar

Is this another one of your twitter jokes?

Maciejasjmj

@ben_lubar said:

Is this another one of your twitter jokes?

If even Ben finds your joke unfunny, you know you screwed up.

tar

Has anyone else noticed that Lennart Poettering's face is the moral and legal equivalent of a large sign which says "please steal all my lunch money and give me a huge wedgie"?

/topic

flabdablet

riking

Ubuntu's Vivid release (15.04) will switch to booting with systemd on Monday

Announce: Vivid will switch to booting with systemd next Monday, brace for impact

tar

OH NO ITS THE END OF THE WORLD LINUX IS RUINED FOREVER EVERYONE WILL ABANDON IT—BUT COULD 2015 BE THE YEAR OF FREEBSD ON THE DESKTOP?

tar

In all seriousness, who really cares about Ubuntu versions which aren't LTS?

Maciejasjmj

@riking said:

on Monday

I read it as "on Mondays" and wondered how silly Ubuntu can get before everyone stops taking these guys seriously.

Apparently, we're not just there yet...

Maciejasjmj

@tar said:

who really cares about Ubuntu versions that aren't LTS?

Desktop Linux users! All, like, two of them.

blakeyrat

Did I link to this Twitter account yet:

https://twitter.com/ShitDevuanSays

Because it's funny as shit and run by a friend of mine.

tar

So, as my friend Miguel de Icaza likes to say, "it's open source, so when it breaks, you get to keep both halves..."

dkf

@tar said:

So, as my friend Miguel de Icaza likes to say, "it's open source, so when it breaks, you get to keep both halves..."

I've known quite a few people use that. I wonder where it came from first. (I know who I got it from, and it wasn't Miguel.)

tar

I was... uh... borrowing someone's blog post there... That's my story and I'm sticking to it.

dkf

I “borrow” stuff all the time. I was just wondering about the provenance trail on that phrase.

HardwareGeek

Google isn't helpful. It doesn't find any results for that exact phrase. Without quotes, it only finds three things that are relevant, and two of those are the same @codinghorror blog post, quoting de Icaza, from earlier this year. The first hit is a question from 2013 on, of all places, english.SE. It doesn't even find de Icaza's own use.

TwelveBaud

No more Upstart, no more Plymouth, no more forked udev, no more PAM...

Why? Systemd does that too. Mandatorily.

dkf

@HardwareGeek said:

Google isn't helpful. It doesn't find any results for that exact phrase. Without quotes, it only finds three things that are relevant, and two of those are the same @codinghorror blog post, quoting de Icaza, from earlier this year. The first hit is a question from 2013 on, of all places, english.SE. It doesn't even find de Icaza's own use.

Well, I knew it from my friend (who works at GE of all places) and that's probably from sometime in 2007 or maybe 2008. (I remember which event it was at, but not exactly which venue, which would let me narrow down to a specific few days in a specific year.) Since I don't think it was original to him, I was just wondering where it came from.

ComputerForumUser

@Maciejasjmj said:

Desktop Linux users! All, like, two of them.

Maybe the other one does. I only use LTS Ubuntu for my desktops.

Actually I just looked, and a fair few other Ubuntu and Steam users use LTS too.

VaelynPhi

@tar said:

In all seriousness, who really cares about Ubuntu versions which aren't LTS?

I occasionally upgrade with the hope that new stuff will work appropriately, like switching between my graphics cards. With 14.04, though, it was a bit of a joke. I can easily switch by relogging, but I lost the use of bumblebee, and CUDA still won't work. ::sigh:: I have to work in Windows for the most part this year anyhow, so I guess it's not much of a loss.

@Maciejasjmj said:

Desktop Linux users! All, like, two of them.

::waves::

@blakeyrat said:

Did I link to this Twitter account yet:

https://twitter.com/ShitDevuanSays

Because it's funny as shit and run by a friend of mine.

I had to google Devuan just to find out what the hell it was. Wouldn't "Ebian" be a much wittier name for that? Hrm... sounds too much like Evian, maybe?

Why can't they just rewrite systemd to respect unix philosophy? It's a tangled web, I'm sure, but it would make more sense than just regressing.

@ComputerForumUser said:

Maybe the other one does. I only use LTS Ubuntu for my desktops.

::waves again::

riking

@VaelynPhi said:

Why can't they just rewrite systemd to respect unix philosophy? It's a tangled web, I'm sure, but it would make more sense than just regressing.

ComputerForumUser:

systemd/src/systemctl/systemctl.c at 5ad096b3f1331b175340129a8c9a5a9d711e5415 · systemd/systemd

The systemd System and Service Manager . Contribute to systemd/systemd development by creating an account on GitHub.

7,435 lines.

tarunik

I've been running with systemd (Debian testing on a desktop) for oh, a couple months or so now, and it seems to function; at least, I have not seen any sparks flying as a result of the transition.

Steve_The_Cynic

@cartman82 said:

- It was deemed that a lot of beloved linux tools were better off as modules of systemd than as separate apps. Goodbye crond, syslog etc... Instead, you get crummy stand-in shivs that advertise systemd replacements every time you run them.

I don't think this word means what you think it does. Or at least I hope you don't mean that the stand-ins are the sort of improvised knives that prisoners make...

(I suspect you meant "shims", but I'd be entertained if you were to correct me and say that you did really mean improvised prison knives.)

accalia

good to know.... of course now all my custom upstart scripts will need to be rewritten....