| View previous topic :: View next topic |
| Author |
Message |
cach0rr0
Moderator
Joined: 13 Nov 2008
Posts: 4117
Location: Houston, Republic of Texas
|
 Posted: Thu Apr 26, 2012 9:27 pm Post subject: |
 |
|
| saellaven wrote: |
I've had syslog and other programs start spewing errors, run away and fill up /var or /tmp and they would have filled up the entire disk if they weren't on separate filesystems. |
logs == old and busted
binary format == new hotness
if only we had someone who understood what an awesome idea it is to swap out plain old text logs with binary blobs...
_________________
Lost configuring your system?
dump lspci -n here | see Pappy's guide | Link Stash |
|
| Back to top |
|
 |
avx
Advocate
Joined: 21 Jun 2004
Posts: 2063
|
| Posted: Thu Apr 26, 2012 9:37 pm Post subject: |
|
|
Can someone tell me, what the purpose of binary logs is supposed to be?
I mean, normal p/t logs are
- human readable
- easily compressed
- easily searched
- subject to work forever
- can be written easily by any application
while binary logs are
- not human readable
- need special tools to be readable
- need special toos/libs to be writeable
- may no longer be readable, when upstream implements a new format and there's no old version lying around.
As for dbus, is that really needed in a normal usecase of a desktop user? Aren't FIFOs and other "old things" good enough for the messaging needs? Is it only, that there are so easily useable APIs/bindings, that dbus took off or where's the great benefit? And if dbus is so great, why are some apps also adding another way of remote control(lastly seen with e17-remote)?
_________________
++++++++++[>+++++++>++++++++++>+++>+<<<<-]>++.>+.+++++++..+++.>++.<<+++++++++++++++.>.+++.------.--------.>+.>. |
|
| Back to top |
|
|
John R. Graham
Administrator
Joined: 08 Mar 2005
Posts: 6434
Location: Somewhere over Atlanta, Georgia
|
| Posted: Thu Apr 26, 2012 10:02 pm Post subject: |
|
|
Please tell me that the binary log comments are in the vein of, "What else will they screw up next?," as opposed to part of some already-in-place plan.
- John
_________________
This space intentionally left blank. |
|
| Back to top |
|
|
avx
Advocate
Joined: 21 Jun 2004
Posts: 2063
|
| Posted: Thu Apr 26, 2012 10:19 pm Post subject: |
|
|
John, did you miss out on that? 
http://lwn.net/Articles/468049/
https://docs.google.com/document/pub?id=1IC9yOXj7j6cdLLxWEBAGRL6wl97tFxgjLUEHIX3MSTs&pli=1
| Quote: | Data fields are compressed in order to save disk space. The net effect is that even though substantially more meta data is logged by the journal than by classic syslog the disk footprint does not immediately reflect that.
The on disk format uses exclusively 64bit LE (little endian) offsets, in order to simplify things and ensure we can store blob data of substantial sizes. No synchronization between log browser tools and journald is necessary, clients which want to browse the journal files can simply mmap() the journal files and use file change notifications for information about updates.
A client library to allow access to the journal files is available, |
| Quote: | My application needs traditional text log files on disk, can I configure journald to generate those?
No, you can’t. If you need this, just run the journal side-by-side with a traditional syslog implementation like rsyslog which can generate this file for you.
Why doesn’t the journal generate traditional log files?
Well, for starters, traditional log files are not indexed, and many key operations very slow with a complexity of O(n). The native journal file format allows O(log(n)) complexity or better for all important operations. For more reasons, refer to the sections above. |
Nuff said? Granted, there are some interesting ideas, but the proposed implementations are giving me shivers.
_________________
++++++++++[>+++++++>++++++++++>+++>+<<<<-]>++.>+.+++++++..+++.>++.<<+++++++++++++++.>.+++.------.--------.>+.>. |
|
| Back to top |
|
|
Dont Panic
Guru
Joined: 20 Jun 2007
Posts: 317
Location: SouthEast U.S.A.
|
| Posted: Thu Apr 26, 2012 10:57 pm Post subject: |
|
|
| When I see some of the local exploits that cause so much commotion, and I compare it to dbus' default capabilities, I have a hard time discerning the difference. |
|
| Back to top |
|
|
cach0rr0
Moderator
Joined: 13 Nov 2008
Posts: 4117
Location: Houston, Republic of Texas
|
| Posted: Fri Apr 27, 2012 4:48 am Post subject: |
|
|
| John R. Graham wrote: | Please tell me that the binary log comments are in the vein of, "What else will they screw up next?," as opposed to part of some already-in-place plan.
- John |
this is the brain that is driving the entirety of Linux' roadmap. Something to keep in mind.
Everything moving forward is being built on this....thing.
The curious one to me, "if you have a better way to do it, please send patches"
Right. So I'm supposed to create patches to unfuck something that, prior to your screwing with it, wasn't fucked. It wouldn't need patches if you didn't screw with it at the behest of the Good Idea Fairy! It works. We like it. Stop screwing with it.
I remember a friend telling me a story a while back - a maintainer for a distro's dmcrypt package. He ended up being dinged for not submitting any patches for "excessively" long - his answer? "There are no patches to submit, because there are no bugs, and it works. If you find any bugs, I will update and commit, but I'm not committing just for the sake of timestamp++"
Familiar territory. Let's rewrite something that works quite well just because, well, fuck it, we haven't rewritten this in a while!
_________________
Lost configuring your system?
dump lspci -n here | see Pappy's guide | Link Stash |
|
| Back to top |
|
|
avx
Advocate
Joined: 21 Jun 2004
Posts: 2063
|
| Posted: Fri Apr 27, 2012 7:00 am Post subject: |
|
|
| cach0rr0 wrote: | | Familiar territory. Let's rewrite something that works quite well just because, well, fuck it, we haven't rewritten this in a while! |
A typical evening in the premises of Lord Poettering, watching a little tv:
Poettering, looking around the room: "Mh, the walls look shabby, might give them a new paintjob... oh screw it, while I'm at it, might as well tear them down and build a new house. It's spring, the birds sing, I can live for a while without the blessings of a working environment."
TV: "ATTENTION, as a result of Lord Poettering playing Jumanji last night..."
P: "How do they know? Ah, forgot, I finally got autoblog-my-life working and included in systemd last night, might have left it on."
TV: "...hundreds of thousands of wild animals are roaming the streets, trampling over anything and killing anyone in there paths. Additionally, the pulse of a new audio system..."
P: "...yeah, Lady Gaga sounds more Gaga with PulseAudio. I'm so awesome, for realzors."
TV: "...have shattered the earth, resulting in the eruption of a super volcano. While magma is flowing down the streets, hurricanes are on their way, resulting from the new differences in air temperature. You're advised not to leave your home!"
P: "Suck it, walls need a paintjob. I'm the Lord, I'll do it and if you're not down with that, suck it ... ehm ... send patches!"
_________________
++++++++++[>+++++++>++++++++++>+++>+<<<<-]>++.>+.+++++++..+++.>++.<<+++++++++++++++.>.+++.------.--------.>+.>. |
|
| Back to top |
|
|
Dr.Willy
Apprentice
Joined: 15 Jul 2007
Posts: 287
Location: NRW, Germany
|
| Posted: Fri Apr 27, 2012 9:39 am Post subject: |
|
|
| Something I've been meaning to ask: Has anyone had a look at dev9? It seems like it didn't get a lot of attention lately, but if people are increasingly upset about udev it might be worth putting some more time into it. |
|
| Back to top |
|
|
avx
Advocate
Joined: 21 Jun 2004
Posts: 2063
|
| Posted: Fri Apr 27, 2012 11:15 am Post subject: |
|
|
http://becquerel.org/dev9 you mean? Depends also on duat and curie, don't know their state.
_________________
++++++++++[>+++++++>++++++++++>+++>+<<<<-]>++.>+.+++++++..+++.>++.<<+++++++++++++++.>.+++.------.--------.>+.>. |
|
| Back to top |
|
|
cach0rr0
Moderator
Joined: 13 Nov 2008
Posts: 4117
Location: Houston, Republic of Texas
|
| Posted: Fri Apr 27, 2012 8:16 pm Post subject: |
|
|
| avx wrote: | | http://becquerel.org/dev9 you mean? Depends also on duat and curie, don't know their state. |
haven't seen mdeininger on since about the last time any of those were updated.
_________________
Lost configuring your system?
dump lspci -n here | see Pappy's guide | Link Stash |
|
| Back to top |
|
|
VoidMage
Advocate
Joined: 14 Oct 2006
Posts: 4810
|
| Posted: Fri Apr 27, 2012 8:28 pm Post subject: |
|
|
Just a small thing: how many people in this thread already know that in the first week of April udev was merged into systemd ?
According to the announcement, it should still be usable without systemd, but do you feel the thrill ? |
|
| Back to top |
|
|
avx
Advocate
Joined: 21 Jun 2004
Posts: 2063
|
| Posted: Fri Apr 27, 2012 9:43 pm Post subject: |
|
|
| VoidMage wrote: | Just a small thing: how many people in this thread already know that in the first week of April udev was merged into systemd ?
According to the announcement, it should still be usable without systemd, but do you feel the thrill ? |
Rather sooner than later, they'll have hard deps on each other, yes, I fear the same. For me, mdev is working quite nicely so far, though.
_________________
++++++++++[>+++++++>++++++++++>+++>+<<<<-]>++.>+.+++++++..+++.>++.<<+++++++++++++++.>.+++.------.--------.>+.>. |
|
| Back to top |
|
|
cach0rr0
Moderator
Joined: 13 Nov 2008
Posts: 4117
Location: Houston, Republic of Texas
|
| Posted: Sat Apr 28, 2012 7:33 am Post subject: |
|
|
| VoidMage wrote: | Just a small thing: how many people in this thread already know that in the first week of April udev was merged into systemd ?
According to the announcement, it should still be usable without systemd, but do you feel the thrill ? |
I'd heard it was going to happen. I had hoped they would realize this is a blindingly stupid idea. But no, I did not realize it had already happened.
I'd love to see a mass migration to mdev as a result. In fact it would be interesting to have a stickied "Get rid of udev" thread that, along with the wiki, semi-officially helps users rid themselves of udev, and do the migration over to mdev.
I will be doing this mdev change in my test lab at work, and shortly after - assuming no significant hiccups - migrate our ~120 Gentoo servers (just ours, not counting customers, a number that should quadruple within the next 6 quarters) to mdev. I think the folks maintaining busybox have a bit more sense than to merge busybox with systemd ^_^
Seriously this systemd/udev idiocy is the first change in the Linux world that has made me itch to leave Linux for something like FreeBSD in the ~15 or so years I've been using it. If linux becomes unusable without upulsedevsystemdaudio I will be gone, and I'll take everyone I can with me - predominately other corps/businesses that are feeding money into linux development, who know our only supported platforms are RHEL and Gentoo, who will follow us whatever decision we decide to make. I'll pay the money out of my own pocket for them to do the migrations if I have to, I refuse to reward stupid with cash.
_________________
Lost configuring your system?
dump lspci -n here | see Pappy's guide | Link Stash |
|
| Back to top |
|
|
genstorm
Advocate
Joined: 05 Apr 2007
Posts: 2241
Location: Austria
|
| Posted: Sat Apr 28, 2012 8:14 am Post subject: |
|
|
It seems to be clear that Debian will not adopt systemd as its default unless it is possible to be run on *BSD, unless Debian change their priorities. Given that, a fork of udev in case of hard-deps on systemd creeping in is likely.
_________________
backend.cpp:92:2: warning: #warning TODO - this error message is about as useful as a cooling unit in the arctic |
|
| Back to top |
|
|
avx
Advocate
Joined: 21 Jun 2004
Posts: 2063
|
| Posted: Sat Apr 28, 2012 8:28 am Post subject: |
|
|
| genstorm wrote: | | It seems to be clear that Debian will not adopt systemd as its default unless it is possible to be run on *BSD, unless Debian change their priorities. Given that, a fork of udev in case of hard-deps on systemd creeping in is likely. |
One of the very few things, I actually do respect about Debian, indeed.
As far as my Google-fu brings me, it seems SUSE and therefore KDE are also in on systemd, that's sad.
_________________
++++++++++[>+++++++>++++++++++>+++>+<<<<-]>++.>+.+++++++..+++.>++.<<+++++++++++++++.>.+++.------.--------.>+.>. |
|
| Back to top |
|
|
cach0rr0
Moderator
Joined: 13 Nov 2008
Posts: 4117
Location: Houston, Republic of Texas
|
| Posted: Sat Apr 28, 2012 9:02 am Post subject: |
|
|
| avx wrote: |
As far as my Google-fu brings me, it seems SUSE and therefore KDE are also in on systemd, that's sad. |
what i dont get is why an operating system whose primary breadwinner is the server platform would make the downright stupid business decision of basing their roadmap exclusively on the desktop, saying to hell with the server folks. I doubt the folks manufacturing the Unimog are going to be enlisting the services of the SMARTCar engineering team any time soon, focusing on how easy it is to park a Unimog on a crowded San Francisco street.
_________________
Lost configuring your system?
dump lspci -n here | see Pappy's guide | Link Stash |
|
| Back to top |
|
|
avx
Advocate
Joined: 21 Jun 2004
Posts: 2063
|
| Posted: Sat Apr 28, 2012 10:18 am Post subject: |
|
|
| cach0rr0 wrote: | | what i dont get is why an operating system whose primary breadwinner is the server platform would make the downright stupid business decision of basing their roadmap exclusively on the desktop |
And I don't understand, how SUSE is binding itself to the mood of Redhat. Granted, the current version of - in this case systemd - is opensource, I'm sure there are ways to allow it's usage only on RH approved systems in the future.
If devs of sane distris (gentoo, debian, arch, slack, ...) would get together, invite the *BSD-guys in and create a really good alternative, that would be great and made me dig deep in my wallet.
_________________
++++++++++[>+++++++>++++++++++>+++>+<<<<-]>++.>+.+++++++..+++.>++.<<+++++++++++++++.>.+++.------.--------.>+.>. |
|
| Back to top |
|
|
genstorm
Advocate
Joined: 05 Apr 2007
Posts: 2241
Location: Austria
|
| Posted: Sat Apr 28, 2012 10:39 am Post subject: |
|
|
| John R. Graham wrote: | | Please tell me that the binary log comments are in the vein of, "What else will they screw up next?," as opposed to part of some already-in-place plan. |
I read it's just optional.
_________________
backend.cpp:92:2: warning: #warning TODO - this error message is about as useful as a cooling unit in the arctic |
|
| Back to top |
|
|
skunk
Guru
Joined: 28 May 2003
Posts: 570
Location: granada, spain
|
| Posted: Sat Apr 28, 2012 10:52 am Post subject: |
|
|
there is one thing that's not clear to me: is just the separate /usr partition affected or even the separate /var partition on server installations?
i've only a couple of old server installations with separate /usr but all have a separate /var...
will a initramfs be mandatory for them too? |
|
| Back to top |
|
|
mv
Advocate
Joined: 20 Apr 2005
Posts: 3140
|
| Posted: Sat Apr 28, 2012 10:58 am Post subject: |
|
|
| genstorm wrote: | | John R. Graham wrote: | | Please tell me that the binary log comments are in the vein of, "What else will they screw up next?," as opposed to part of some already-in-place plan. |
I read it's just optional. |
As optional as all the *-kit nonsense was claimed to be.
The good side is that everybody who was not sure that this guy lost any sense of reality and has no idea how a concept for an operating system should look like now knows better.
Concerninig "send patches": For a daemon running with root permissions (policykit insanity) the only reasonable patch is to not run such a nonsense daemon in the first place, i.e. rm all Poettering stuff. Unfortunately, this requires to send patches to a lot of upstreams (and to convince them) to avoid this stupidity. This is almost impossible if these people are paid by the same people who could not see that Poetterings suggestion were all completely broken by design. |
|
| Back to top |
|
|
genstorm
Advocate
Joined: 05 Apr 2007
Posts: 2241
Location: Austria
|
| Posted: Sat Apr 28, 2012 10:59 am Post subject: |
|
|
The *kit nonsense was getting depended on by more and more userspace packages, I don't see how that should be the same for the system log option. I might be wrong, though.
@skunk: >=udev-180 expects /usr to be mounted, regardless of systemd. Nothing to be read about /var, I guess the discussion would have evolved into a major shitstorm if else.
| avx wrote: | | If devs of sane distris (gentoo, debian, arch, slack, ...) would get together, invite the *BSD-guys in and create a really good alternative, that would be great and made me dig deep in my wallet. |
The Debian mailing list discussion was going into the direction of cautious systemd approval, but rather bringing some necessary changes to the *BSD kernel (such as cgroups functionality) and perhaps maintaining a systemd-patchset (if not accepted upstream) to being able to use it across kernels.
_________________
backend.cpp:92:2: warning: #warning TODO - this error message is about as useful as a cooling unit in the arctic
Last edited by genstorm on Sat Apr 28, 2012 11:24 am; edited 1 time in total |
|
| Back to top |
|
|
skunk
Guru
Joined: 28 May 2003
Posts: 570
Location: granada, spain
|
| Posted: Sat Apr 28, 2012 11:21 am Post subject: |
|
|
| genstorm wrote: | | @skunk: >=udev-180 expects /usr to be mounted, regardless of systemd. Nothing to be read about /var, I guess the discussion would have evolved into a major shitstorm if else. |
ok, thank you...
i'll mask >=udev-180 on these servers hoping older udev won't get removed from portage and future packages won't ask newer udev as dependency...
in the mean time hopefully somebody will realize this mess must be fixed, else there already is a mdev wiki in place...
i can sleep soundly again |
|
| Back to top |
|
|
genstorm
Advocate
Joined: 05 Apr 2007
Posts: 2241
Location: Austria
|
| Posted: Sat Apr 28, 2012 11:28 am Post subject: |
|
|
I did the same on my ~arch system until things get ironed/worked out, I don't like initramfs.
However, there are already newer package versions requiring >=udev-180:
| Code: | # cat /etc/portage/package.mask/playsafe
#ffs!!
>sys-fs/udev-180
>sys-auth/consolekit-0.4.5-r2
>sys-apps/pciutils-3.1.9-r1
>sys-apps/usbutils-005
>sys-apps/lshw-02.16b |
_________________
backend.cpp:92:2: warning: #warning TODO - this error message is about as useful as a cooling unit in the arctic |
|
| Back to top |
|
|
avx
Advocate
Joined: 21 Jun 2004
Posts: 2063
|
| Posted: Sat Apr 28, 2012 11:39 am Post subject: |
|
|
@genstorm, what do cgroups actually have to do with init? AFAIK, they've got to be setup by userspace anyway, so I fail to see a need in depending on it - while it's currently optional. So why changes to the BSD kernel, do they even want it?
I read somewhere, that it's not only effecting /usr, but also /var, don't know if that is now or in the future.
_________________
++++++++++[>+++++++>++++++++++>+++>+<<<<-]>++.>+.+++++++..+++.>++.<<+++++++++++++++.>.+++.------.--------.>+.>. |
|
| Back to top |
|
|
genstorm
Advocate
Joined: 05 Apr 2007
Posts: 2241
Location: Austria
|
| Posted: Sat Apr 28, 2012 11:51 am Post subject: |
|
|
On the Debian mailing list, this was mentioned as being 'really cool', dunno 
From somewhere else:
| Quote: | | cgroups were introduced into the Linux kernel a few years ago, and they are an interesting mechanism for allocating and limiting kernel resources. In systemd, cgroups are used to corral and manage processes. When new processes are spawned they become members of the parent's cgroup. The cgroup is named for the service it belongs to, and services cannot escape from their cgroups so you always know what service they belong to. When you need to kill a service you can kill the cgroup, and snag all of its processes in one swoop instead of playing find-the-fork-or-renamed-process. Another way to view the process hierarchy is with the system-cgls command, as shown in Figure 2. |
| avx wrote: | | I read somewhere, that it's not only effecting /usr, but also /var, don't know if that is now or in the future. |
http://comments.gmane.org/gmane.linux.hotplug.devel/17167
EDIT: Hmm ok, steveL wrote in his thread about /var being mandatory. ORLY? Everyone else on the webs was just talking about /usr...
EDIT2: And if we can believe L.P., it is not true.
_________________
backend.cpp:92:2: warning: #warning TODO - this error message is about as useful as a cooling unit in the arctic |
|
| Back to top |
|
|
|
|
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum
|
|
Portage & Programming
