| DistroWatch Weekly
|Linux Foundation Training
|Reader Comments • Jump to last comment
1 • reports of XFS's demise are exaggerated! (by greenpossum on 2012-11-05 10:42:52 GMT from Australia) |
>Now that SGI is no longer with us, all development work on XFS has ceased.
Not so! Dave Chinner has been hard at work and large improvements were reported earlier this year, see the first link.
2 • Tips and Tricks (or about JFS) (by Anonymous on 2012-11-05 11:00:42 GMT from Brazil)
One may use JFS, of course. I've did that for some time, based on this old article (http://www.debian-administration.org/articles/388), from which I concluded JFS was the filesystem which consumes CPU the minimum. That time, I had an old machine which very little CPU resources, and using JFS seemed to make it work better, smoother.
However, I noticed that when commands like "find / ..." run in a JFS filesystem, everything else is *very* sacrificed. It seems to be a result of some options assumed in the heart of JFS.
The same article put XFS in a gold place, if one has sufficient resources, which is reasonable for current hardware.
3 • Lost Hard Drive! What am I going to do? (by JS on 2012-11-05 11:39:12 GMT from United States)
I'll tell you what I did, I installed Linux on a spare Sandisk I have, and got my computer running again with hardly any noticeable difference. You can't save yourself that way with Windows. With Linux, you don't need a hard drive, unless you are a heavy gamer or need space beyond that of a Sandisk. My Sandisk is only 8 gigs, but that is all I need.
4 • What about btrfs? (by Alok on 2012-11-05 11:44:40 GMT from India)
I thought Btrfs merited a mention. Its fast and supports journals and snapshots and what not.
5 • Tails and Ext4 (by Barnabyh on 2012-11-05 11:49:12 GMT from United Kingdom)
Thanks to this review I will try Tails soon for my privacy needs, had Tor previously on my now old Slackware 13.37 box with most traffic routed through it, but the hd crashed. It seems a bit too old now to roll it back onto the replacement.
Which brings me to the second topic. I never trusted ext4 fully, although I can't provide any links I seem to recall postings about corruption and data disappearing very early on, and had a few instances myself although it's ghard to prove for me as the version of PCManFM which I was using at the time was also said to cause data loss and data corruption.
Anyway, funny enough both partitions that developed unrecoverable faults were using ext4, even a low level reformat does not help once you got damaged blocks. It may be related or not, but it became more problematic after using my own 3.4 kernel for a while, and from then on I frequently got I/O errors, 100% cpu spike and freezes due to being unable to read the fs.
So I'm back to reliable, fast, good old Xfs as well, ext2 for the boot partition and Puppy. As pointed out by Robert, and this goes for Puppy and derivatives too, most need xfsprogs added which is a quick affair indeed.
Also, don't forget to change fstab for the reformatted partitions.
@3 Very nice indeed.
6 • Claws/Gmail and Debian default desktop. (by iki on 2012-11-05 12:06:59 GMT from Brazil)
Claws is light and works great but when using with Gmail a few months ago it flooded the message thread with partial messages (like if it was saving drafts as full sent messages) so I stick with Thunderbird.
I think Debian should go with Xfce if it was 4.10 but since it seems they are stuck with 4.8 ...
7 • ext4 data corruption bug (by megadriver on 2012-11-05 12:24:10 GMT from Spain)
Oh, no. Not the ext4 thing again.
I haven't posted here in a while, but today duty calls! :)
Everybody should read this:
The tone of today's article sounds dangerously close to FUD, IMHO.
I'm running Arch testing, and I've had zero data loss with ext4. Even after some recent "dirty" reboots due to blackouts, it's been rock solid here. The experience of others may be different, of course, but this particular bug is definitively not the culprit, unless you do some really weird stuff with your ext4 filesystem, as explained in the link above.
All modern filesystems are complicated things, so they are expected to have bugs. I doubt current ext4 is particularly worse than others that regard. If you have important data that you can't afford to lose, no matter the filesystem you use, regular backups should be made. It's common sense!
On a lighter note, I must say the concept of the sposkpat distro highly amuses me. :)
8 • EFF sugests diff desktop (Ubuntu) (by John Dough on 2012-11-05 12:42:09 GMT from Canada)
Yep, it's called Kubuntu, polished and it works. I have a little rule of my own, if I want a certain DE, I go with a distro that has that DE as the default, everything is more likely to work.
9 • Filing Systems (by Looking for basics on 2012-11-05 12:52:12 GMT from Canada)
Have been using Reiserfs version 3 for years now without any trouble. Find it faster than ext and possibly more efficient on using disk space (going by memory on that - can't find my install notes from years ago doing the comparison). In some cases, with tons of small database files, had to format with extra inodes on ext systems. Plus there was the irritating bootup delay at some intervals while doing a fschk (seem to recall from the manpage there are options to control when and at what intervals this is performed). With reiserfs, this is all done in the background, no issues with inodes encountered so far, and have never had any trouble with it.
Never got around to playing with reiserfs version 4. It was during it's develpment phase that the whole scandal over Hans Reiser's conviction erupted, delaying further work on that filing system. Would be curious to know if others have played around with reiserfs 4, what their experiences were, and how performance ranked with the other filing systems.
Final question, never answered from previous post a few weeks ago, concerns status and performance of a suitable recovery tool for btrfs. Btrfs is supposed to be almost ready for prime time, but would you trust your data to a system without a recovery tool or a proven track record?
10 • ext4 (by Jesse on 2012-11-05 13:16:45 GMT from Canada)
>> " I never trusted ext4 fully, although I can't provide any links I seem to recall postings about corruption and data disappearing very early on"
Yes, ext4 did have some problems very early on in its development. As almost all file systems do until they've had a wide range of testing under a wide variety of loads. This is why I usually recommend people use a file system that isn't the latest and greatest. For example, all my machines still use ext3 instead of ext4 or brtfs. When new releases of ZFS come out I hang back a version or two as I figure, with file systems, if it's not broke, don't fix it.
>> "The tone of today's article sounds dangerously close to FUD, IMHO. I'm running Arch testing, and I've had zero data loss with ext4."
Most people won't have data loss, even with this bug. That doesn't mean the bug doesn't exist. It's there, it's being worked on and, for those few people it affects, it's quite serious.
11 • Is the EXT4 problem exaggerated? (by David Smith on 2012-11-05 14:33:19 GMT from Canada)
I use ext4 exclusively for several linux installations both at home and work, have done for quite a few years, so like many I was initially quite concerned about the report of a rather devastating-sounding bug manifesting in latest kernels. However further reading indicates the bug can only be summoned through uncommon (and frankly, pretty weird-sounding) use-cases.
If you take a spin through current Fedora and Ubuntu forums, you will encounter some discussions about the bug, but virtually no reports of it actually hitting users.
So based on that, yes this is being exaggerated. And it's unclear to me if the underlying issue is with ext4 per se, or with newer kernels handling of it.
12 • EXT4 corruption (by claydoh on 2012-11-05 15:05:55 GMT from United States)
This is all anecdotal, just like anyon'es account of everything beeing peechy keen, but I have seen a small handful of users report similar hard drive issues where suddenly errors pop up up about foo.conf being unwritable: check permissions, followed by an automatic fsck upon reboot, which fixes it for a while. I myself thought I had 2 bad drives due to these errors because of the frequency of this happening. But surprisingly they are quite fine now, running as external usb drives. My internal drive is now a small ssd with no EXT4 issues at all.
While I think that the particular bug mentioned is probably not what is effecting these users, it may very well be a different bug. But alas, how do we reproduce it and how do we discover where or what it is? That is the tough part.
13 • XFS and that old recovery problem chestnut (by gord on 2012-11-05 16:25:21 GMT from United Kingdom)
the thing that stops me using XFS is that the only recovery tool I know of is a commercial offering. I could be unlucky, but I've lost about 5 XFS volumes over the years, all due to "sudden volume dissaperance" (partitions there, filesystems don;t get recognised by xfs_repair or _dump. Of those, 4 of them could be about 90% recovered using the commercial recovery suite, most of those with filename intact but not directory heirarchy.
Does anyone know of a FOSS file-level (not block-level like dd_rescue et al) that can recovery files from borked XFS volumes?
14 • ext4 corruption rider (by gord (again) on 2012-11-05 16:28:23 GMT from United Kingdom)
it should be pointed out again that the recent corruptions are just that: recent, and afflict only recent kernels.
Don't run bleeding edge if you love your data. In fact, I personally run middle-ground ZFS, because I love my data ;)
15 • Tails review (by Fabio on 2012-11-05 16:50:14 GMT from United States)
Very nice review Jesse, few issues back a Linux magazine in the UK (and where else anyway? in the USA aint no Linux magazines..no comment..) was giving out a dvd with Tails
The only thing I don't like about Tails is that it doesn't fit into a CD. Not sure how important is GIMP and Inkscape are, these are application for pros, it's like why is AutoCAD not installed then? I'm sure Tails users also have their own default PC with all their sofware installed and when they want to be naughty on the web they use Tails.
16 • Alt file systems (by Patrick on 2012-11-05 16:54:57 GMT from United States)
I used to keep most of my data on an XFS file system (now I think about it... I probably still do! :-)), while my system disk has had ReiserFS or XFS, or ext4 more recently. On my SheevaPlug I've been running JFS due to its reputation for low CPU usage. I had some reservations about JFS, especially because it was hard to find anyone recommending it, but I've had zero problems with either XFS or JFS in the years I've been using them. And JFS definitely seems to run very smoothly on the little SheevaPlug CPU.
17 • ext4 bug is a non-issue for 99.99% (by Scott Dowdle on 2012-11-05 19:26:43 GMT from United States)
Linux Weekly News had a good write up on how the ext4 issue was overblown. Unfortunately that article is less than a week old and currently subscriber only... but it should be freely available starting Thursday, Nov. 8th (https://lwn.net/Articles/521803/). Yeah there is a bug... but there are bugs everywhere... what's important is what it takes to hit the bug and how many people it affects. According to everything I've read, one has to really go out of their way by changing default options and so far as I know, none of the distros do that by default. That begs the question, how could Robert Story run into the issue 2 times without doing (one would assume) the manual steps necessary to create the conditions that would trigger the bug? The answer is that he couldn't have. Obviously he ran into some other issue... but what issue was it? That would be interesting to know... because whatever it was, wasn't this bug.
I totally agree that his article as it is now, is complete and utter FUD. I'm not saying he wrote it as FUD on purpose but that is what it is.
18 • Bugs, bugs everywhere (by Jesse on 2012-11-05 19:43:49 GMT from Canada)
>> "but there are bugs everywhere... what's important is what it takes to hit the bug and how many people it affects."
There is a third factor I feel needs to be in that list: what happens when the bug is triggered? A bug which causes my instant messenger client to crash is a minor annoyance. A bug which causes the data on my hard drive is become corrupted or disappear is a pretty serious bug.
>> "Obviously he ran into some other issue... but what issue was it? That would be interesting to know... because whatever it was, wasn't this bug. I totally agree that his article as it is now, is complete and utter FUD."
If you are right and he ran into a separate bug that causes the data on a hard drive to become corrupted, wouldn't that indicated the problem is much _more_ serious rather than less serious? The idea that there is a second, yet unknown bug which affects multiple distributions and causes data loss sounds like a pretty big deal to me. I'm hoping you're wrong and it is the same bug and, once it's fixed, the issue can be laid to rest.
19 • ext4 bug fixed? (by Nick on 2012-11-05 20:23:17 GMT from United Kingdom)
It seems that the infamous bug is fixed now (in kernel 3.6.6).
See commit ffb5387e85d528fb6d0d924abfa3fbf0fc484071 at the bottom of: http://www.kernel.org/pub/linux/kernel/v3.0/ChangeLog-3.6.6
Or am I missing something?
20 • ext4 bug fixed? (by Nick on 2012-11-05 20:26:40 GMT from United Kingdom)
I mean commit dab43b73a7c7317f941c1314e9a77374ba8999ee
21 • Speed up ext3? (by uz64 on 2012-11-05 20:41:26 GMT from United States)
Yeah, ext3 is slow. But you can mount it with the ext4 driver for some increased performance. Hey, the bug supposedly only affects ext4 file systems, right? I seriously doubt that mounting an ext3 file system with the ext4 driver will trigger this bug, but you never know. To be sure you're on the safe side you could also mount it with the ext3 driver at the expense of speed, but when it comes to potentially devastating data loss, that's a sacrifice that may be worth it.
ext3 does not do delayed allocation by default unlike ext4, so at least you'll know for sure that when you copy a file over, it is transferred immediately, no wait... so you could pull the plug after the drive is done grinding and probably suffer no major loss. Which is more than can be said about ext4, which waits until it feels like it to actually write data to disk, so you may end up with a file full of zeros or something unless you enter the "sync" command first. All in an effort to speed the file system up (apparently it's worked, because you're complaining about its predecessor's speed).
XFS is good, I agree--I used to use it in my first years on Linux, and I liked the fact that it comes with a defragmentation program. JFS is also very nice; no defragmentation program, maybe not quite as fast as XFS overall, but lighter on resources.
22 • Another bug? (by Scott Dowdle on 2012-11-05 21:40:17 GMT from United States)
@18 - Jesse
That's just it. I find it hard to believe that a second bug exists... like Robert reports... because if it did, and were so easy to trigger... there would be all kinds of people having issues. The fact that there aren't should indicate something else. What that something else is, I don't know... but yeah, I'd like someone to look into it. All we have so far is one guy (unless there are others?) who says he easily triggered the bug in question twice from default installs... which, as all technical reports on the issue show, ranges from extremely unlikely to impossible. You seem to want to claim that that is proof that ext4 is even more flawed that we thought. I'm not trying to turn this into a pissing match. I just want level headed, factual reporting... not speculation based on anecdotal evidence. Did he report the issue to the distros? Do they already have bug reports on this? You'd think the distros would be all over it... if it were... indeed... real.
Regarding XFS, I too hear it is a fine filesystem... but like some other commenter here mentioned... I've heard reports that if it ever has issues, it can be problematic to repair. I have no real evidence of that and it would be wrong for me to make negative claims about XFS as a result. I feel somewhat dirty even writing that.
23 • Linux, Get Over It (by AndySysops on 2012-11-05 23:22:05 GMT from United States)
I see ext4 as Linux's not-invented-here syndrome playing catch-up with XFS, which long ago had journals and extents, the "new and improved" bits in ext4.
You can run ext4 sans journal (and so this bug). On flash drives, you should:
mkfs.ext4 -v -t ext4 -i 8192 -O ^has_journal -L MyDiskName /dev/disk/by-id/...
Nowadays Tso himself wants Btrfs. Then there is ZFS. Seems that ext4 is a lost bastard child. The zealotry of ext4 advocates claiming "anecdotal" "FUD" puzzles me. Linux has an ego.
I love the phony h-online party lines on this bug. There's the usual cop-out; all software has bugs. Then comes this vagueness: "XFS...has become stuck on the sidelines." What does "stuck" mean? XFS works better than ext* and has for years. If Linux kernel dweebs would promote it, and stop the not-invented-here attitudes, then people would use XFS more.
ext3 data loss first drove me to XFS. I've used XFS for many years with none whatsoever, lightning crashes and all. I even use XFS-formatted TrueCrypt volumes.
Linux practices subtle vendor lock-in almost as well as Apple and Microsoft. The result: the single most cross-platform filesystem is Microsoft FAT.
Linux distros should all ship XFS/NTFS/FAT/HFS+ tools stock, and compile kernels with full support. Linux (zealotry) says one can recompile a kernel, or modprobe or apt-get . Yeah, right! Explain those tricks to your auntie or beer buddy. I sometimes wonder if Linux really wants to win the desktop, or just prove its 31it3n355.
My own philosophy avoids all Linux-native journals. I will use ext4 sans journal. If I want a journal, I will use XFS.
I've not yet tried ZFS, but only because Linux vendor lock-in / license lock-out makes it harder to use than XFS.
Linux needs to get over itself. Some systemd critics say it's really becoming the "RedHat Linux kernel" now - with special systemd tweaks which only RedHat uses - yet another corporate product.
I've personally felt that potty-mouth Linus prefers his monolithic kernel not because it's technically superior to microkernels, but because it means full political control. A technically superior microkernel architecture would mean yielding power to outsiders. So goes so much of this whole Linux show, just like corporate PR, political / marketing / lock-in gimmicks masquerading as "technical improvements."
Good riddance, ext4.
24 • NIH and lock-in (by Scott Dowdle on 2012-11-06 01:08:11 GMT from United States)
@23 - AndySysops
I see very little support for your claims regarding NIH and lock-in.
Regarding Red Hat... thanks for filling us all in... and exposing Red Hat as the secret open source / free software developer secretly locking us all in... well those of us using systemd. They *TOTALLY* had me fooled with that 10-12% of total kernel development each kernel release... making them the #1 company sponsor... oh... for years now. Obviously they do that, as well as all of the other development on things like libvirt, guestfs, libreoffice, gcc, glibc, gnome, x11, etc... and all of that stuff... just to draw us into their nefarious plan.
If there were only stats to show the deployment stats for the various filesystems on Linux... I'm so sure XFS would win. (cough) But seriously, you like XFS and it works for you, great! More power too you. One feature I like is how XFS supports really large filesystems. ext2/3/4 has historically lagged behind in that and some other areas... but for general purpose use cases, it ain't so bad.
25 • second EXT4 bug (by Cnadide on 2012-11-06 01:50:19 GMT from Taiwan)
There Might Be Another EXT4 Corruption Bug
Posted by Michael Larabel on November 01, 2012
It was only days ago that an EXT4 file-system corruption bug affected the stable Linux kernel, which was finally patched yesterday. Now though it looks like there may be another EXT4 corruption bug affecting the stable kernel.
(see above link for more)
26 • Re. 23 (by uz64 on 2012-11-06 02:49:53 GMT from United States)
"I see ext4 as Linux's not-invented-here syndrome playing catch-up with XFS, which long ago had journals and extents, the "new and improved" bits in ext4."
Well, you must not know that XFS was originally released by SGI for IRIX in 1994, and until 2000 it was not even released under the GPL. In other words, even if people did want to use the file system in Linux... they couldn't, until SGI re-licensed and began porting it themselves. By this time Linux already had its own native file system, ext2, and ext3 was already on its path to coming to existence and replacing it. The whole point of ext4 was, unsurprisingly, meant to provide the same important feature to Linux users that ext3 did: full backwards compatibility with its predecessor (until you enable extents, at least). Hate to break it to you, but XFS--as good as it is--did not and could not do this.
Don't like ext4? Well, don't use it, there are many alternatives in a typical Linux distribution installer. Don't want other people to use ext4? Then why don't you make your own distribution that is XFS-exclusive and try to get people to use it, or just suggest the people you know to use XFS?
"If Linux kernel dweebs would promote it, and stop the not-invented-here attitudes, then people would use XFS more."
Really? It's the job of the Linux kernel guys to promote every obscure feature feature now? I could've swore it was up to the distribution maintainers, who can choose to compile in any feature they want, or leave it out. And in the case of XFS... it is not only readable in almost every distribution, but many also even offer XFS as an option during the install process.
"ext3 data loss first drove me to XFS. I've used XFS for many years with none whatsoever, lightning crashes and all. I even use XFS-formatted TrueCrypt volumes."
ext3 data loss? Never experienced it. XFS data loss? Never experienced that either, although technically due to the way XFS handles disk writes with delayed allocation (just like ext4 introduced), you'd be more likely to end up with a file of all zeros if the power goes out when a file is waiting to be written to an XFS file system.
"Linux practices subtle vendor lock-in almost as well as Apple and Microsoft."
Oh, sure they do. Once again, just use another god damn file system. Hell, use FAT32 or NTFS for /home if you really want. Just don't expect permissions to work properly when using a non-UNIX file system.
"The result: the single most cross-platform filesystem is Microsoft FAT."
Unfortunately this is true, but it is not due to the nonsense you're blabbering on about. It's due to portable device manufacturers (digital camera, portable digital audio player, GPS, hell even Android-based cell phone) insisting on using FAT due to its extreme simplicity to implement and refusal to support anything else. No joke, I have an Android phone and it will NOT read a microSD card that is not formatted in FAT... not even ext2 works on it. And it runs... *gasp*... Linux. Can't get any more native than that, but manufacturers just don't want to support more than FAT. Too bad--because we're the ones buying their products and effectively paying them back for their R&D in the first place.
"Linux distros should all ship XFS/NTFS/FAT/HFS+ tools stock, and compile kernels with full support."
You know what? In a perfect world, you're 100% right. But how much you wanna bet Microsoft will never take those steps with Windows, allowing UFS/UFS2, ext2/3/4, JFS, XFS, ZFS, etc. right upon installation in their OS? So the reality is, it goes both ways. If you really want to encourage people who don't even know WTF they're doing to be granted full R/W access in Linux to NTFS, a closed-spec proprietary that has had to be reverse-engineered just to get running and Microsoft can change at any time... well, I'll just say, you must not have thought that idea through too well.
That said, most distros' installers I've come across allow you to mount partitions (yes, even those with NTFS on them) during the system installation. I tend to avoid doing this myself though and use fstab when I first boot the machine, because I'd rather not risk screwing something up or accidentally formatting a partition that shouldn't be. And I have very little use for NTFS these days, maybe a bit of read-only use every once in a while, but it seems that many (most?) distros come with NTFS-3G already installed... isn't the reverse of that what you're complaining about? That distros don't come with NTFS support, and that they should? So what's the point?
"Linux (zealotry) says one can recompile a kernel, or modprobe or apt-get . Yeah, right! Explain those tricks to your auntie or beer buddy. I sometimes wonder if Linux really wants to win the desktop, or just prove its 31it3n355."
Wow. If you really think that typing something like "sudo apt-get install jfsutils xfsutils ntfsprogs ntfs-3g" is too much voodoo for someone to be typing into what they'll probably just call a "black box," then maybe you could try teaching them the GUI frontend to their distribution's package manager? Or, you know... just do all that scary voodoo black magic yourself? How about you actually help your auntie and beer buddy, just like you probably are now as they sit in front of a Windows machine, instead of whining on forums that you shouldn't have to help users, that everything should just work?
I always thought it was highly ironic how Windows always seemed to get the free pass, and Linux--no matter how many strides it makes toward usability to the technically-braindead masses--is always targeted by people complaining that everything should just work. Hint: Even in the Windows world, that is just NOT THE CASE. My own latest experiences of this come from none other than... drumroll... Windows 8.
"Linux needs to get over itself."
I'm starting to think someone else needs to get over them self...
"Good riddance, ext4."
Okay, so if you have in fact decided to use some common sense and make up your own mind not to use ext4, then what the hell was that rant for?
27 • Re. 25: Debunked. (by uz64 on 2012-11-06 03:21:43 GMT from United States)
No, there's not another known ext4 file system corruption bug. From Theodore Ts'o himself:
It really does sound like Phoronix just wants ad revenue. Ubuntu is the culprit here, not ext4.
28 • Claws Mail (by mjjzf on 2012-11-06 06:10:25 GMT from Denmark)
Claws Mail remains my favorite - I wrote a little piece a while ago for making it a bit prettier: http://writtenandread.net/claws-mail-setup/
29 • ext4 bug issue (by claudecat on 2012-11-06 07:18:18 GMT from United States)
This issue has been blown way out of proportion. I'm surprised that the author was bitten by this bug in the course of doing normal distro installs, as this would appear (from all reports) to be unlikely/impossible unless certain tweaking were done and unless using very specific kernel versions. I'd be more inclined to suspect hardware issues than this bug being encountered with this degree of regularity under these circumstances.
The piece here on DW seems to lend more credence to an issue that is already solved (patch has been out for almost a week now and is being backported to all affected kernels) than it deserves. I won't go so far as to call it FUD, but it is a non-issue for the vast (and I do mean VAST) majority of users.
30 • @23 (by lusr on 2012-11-06 15:43:10 GMT from United States)
I think your computer science teachers are still teaching you from books written in the '80s, when the word "micro-kernel" was associated with a future utopia. -Theo de Raadt
31 • ext4 and microkernel (by GODhack on 2012-11-06 21:09:06 GMT from Lithuania)
I know some micro-kernel fans they run Minix and do not try to tell them that Linux is better. I understand them. But guy suggesting Linux to become Minix was funny one.
Those who suggest XFS to Linux mainstream should use this as start point:
Happy ext4 user. ;)
32 • #31 (by zykoda on 2012-11-07 07:54:09 GMT from United Kingdom)
Anything later than 2009? This applies to GRUB legacy?
33 • #31 (by GODhack on 2012-11-07 15:53:21 GMT from Lithuania)
~ same GRUB 2 or 1. http://archlinux.2023198.n4.nabble.com/grub2-and-xfs-seems-not-to-work-td2038433.html etc. Of course you can use ext2 for /boot xfs for / etc.
34 • File systems are a good place to be conservative (by Gnobuddy on 2012-11-07 20:59:09 GMT from United States)
Whatever small performance gains the "new and improved" file systems promised weren't worth the risk of data loss to me, which is why I've stuck with ext3 on /home on all my Linux boxes, and ext2 for / as well as for any USB keys I plan to use only on my own computers (all of which run Linux).
The biggest performance gains in file I/O operations seem to come from increases in drive spindle rotation speed and/or switching to SSD's, not from changes to the file system.
35 • NOT ext4 and absolutely NOT Reiser! (by YaYo on 2012-11-07 23:17:17 GMT from United States)
If you're now looking for another filesystem to use other than ext4 then JFS may be something to seriously consider. I particularly perked up when I saw that the latest iteration of JFS now supports TRIM (as of 2012). [ http://en.wikipedia.org/wiki/JFS_(file_system) ] That means SSHD's should now be better supported. It also suggests that IBM does indeed still have a hand in the ongoing development. But if you use JFS then you should probably also be aware of the two different versions (and possibly a need for a more beefy CPU)!
JFS1 appears to really only be intended for AIX systems. JFS2, which is referred to as simply as JFS, is more for the Linux crowd. It's not likely that anyone will accidentally install JFS1 since that has more or less been abandoned in most Linux circles. But it may be possible if you use a really old or off beat distro to prepare your system with. Therefore, be aware that JFS2 is the JFS filesystem that you probably want to use if you're going to prepare a standard workstation. (And just like XFS needing xfsprogs, you'll also need the jfsutils to prepare with.)
Robert Storey's suggestion (author of the article "Ext4 Data Corruption Bug and Solution") to possibly consider the ReiserFS also seemed like a good alternative. And I might have wanted to try it. But!!! I am not exactly at ease with the idea of the developer Mr. Hans Reiser being a convicted murderer! Granted, Reiser was convicted of a crime of passion by killing his Russian mail order wife. But to me, that just goes to show how neither of them had very good judgement (although mostly him). I could go on but I think my objections should also be somewhat obvious when considering the potential coercion factor that Mr. Reiser may be under from other inmates while he serves time in the California/Federal penal system. Therefore, trusting the ReiserFS filesystem would sort of like be trusting Bernie Madoff (of the infamous Madoff 100-billion-dollar pyramid investment scandal) to manage my books or even have access to my bank account. It's just NOT gonna happen.
36 • re 35) "Not ext4 and absolutely NOT Reiser!" (by Looking for basics on 2012-11-08 00:35:37 GMT from Canada)
Just curious; if Isaac Newton had hypothetically been convicted of murder, would you deny the existence of Gravity, convinced that apples must always fall UP?
So far as I can recall, Reiser had the support of a whole development team and they continued development of reiserfs version 4 after he was imprisoned for life. I suppose it is possible he may still offer some support in an advisory role, but most likely he is completely out of the picture now. While caution is to be commended, especially with respect to computer security, it would be nice to have some facts, or at least some references before giving a wholesale condemnation.
As for potential for troublemaking; conside the NT filing system (probably one of the most widely used systems in use at the moment): It is possible to hide data in streams within the filing system itself where no one could find it short of resorting to forensic techniques (eg - doing a direct byte analysis of the data on the hard drive).
The upshot is, ALL filing systems go through a testing and gradual rollout process. This appears to be where btrfs is at right now; the last holdup to widespread deployment being a proven recovery tool. Until we get there, rieserfs (well, at least version3) MAY be another alternative for some users....
37 • reiserfs (by DrCR on 2012-11-08 03:46:21 GMT from United States)
I always use reiserfs for OS. XFS and JFS for data storage. (I'm using JFS at the moment though.)
Reiser's conviction through some ice on Reiser4, was well as unmet hype from what I recall, but reiserfs was already well established by that point.
38 • Tor from inside China (by Terence on 2012-11-08 23:14:48 GMT from China)
I read the review of TAILS and decided to download it. As I live in China, I need a way to connect to the net without the great firewall censoring what sites I decide to visit (I also have a VPN, but lately it has proven impossible to connect to). Unfortunately, TOR seems to have great difficulty as well. Any suggestions on how to use TOR effectively from within China? I am a total newb with regards to using it.
39 • @38 Tails & Terence (by greg on 2012-11-09 08:11:13 GMT from Slovenia)
Have you tried Liberte linux? it seem TOR is configured preety much automagically there...
40 • RE: 27 - 34 -36 - 38 (by Landor on 2012-11-09 16:23:22 GMT from Canada)
How do you know that Ubuntu is the problem? I'm not an Ubuntu fanboi either.
I have to agree completely. If you want problem free then use EXT2 and EXT3. The sanity behind your choice is obvious. A lot of people could easily say, 'But the chances are", and yet the chances exist. As you do, I personally prefer something far more robust when it comes to my data, and it's EXT2 and EXT3.
Well said. I've never been vocal about the filesystem myself but I hold the exact same opinion and refuse to even mention it by name.. In a community where a large number of the members are here include philosophical reasons it's a given that we would draw a definitive line. I'm quite sure most of those that denounce such reasoning would protesting against any applications created by, or linked to someone in some manner that created massive atrocities. The number shouldn't be an issue here.
Here's a link to the Tor Project's website, and its documentation page.
Keep your stick on the ice...
41 • correction (by Landor on 2012-11-09 17:17:39 GMT from Canada)
RE:36 should be RE: 35
Keep your stick on the ice...
42 • I don't use JFS due to bad experiences with it (by b_brain on 2012-11-09 17:36:23 GMT from United States)
A number of years ago I had whole directories of data spontaneously disappear, for no apparently reason, from JFS partitions. After the third such occurrence, I switched away from JFS without ever looking back.
I've also used ext2, ext3, ext4, XFS and Reiserfs as well as a number of Unix file systems without having any such problems.
JFS is supposed to be fast and a light resource user, but that has never been enough to make me want to risk my data on it again.
43 • #38 great firewall (by jondo on 2012-11-10 02:03:56 GMT from Australia)
you could try jondo livecd
but look on the bright side, the great firewall could be helping to keep you safe from cybertrolls :)
44 • openSUSE? (by Mac on 2012-11-10 02:23:45 GMT from United States)
I hate to be this way but people here I like to read have said suse in the past. Tried for 3hrs to get the clock to default to cst hay and if you visit the website they think root access is not good. This is what I like about linux is doing it my way and hay openSUSE 12.2 is a joke IMHO. Long live debain.
Have fun Mack
45 • Slax 7 development release (by RollMeAway on 2012-11-10 18:51:56 GMT from United States)
Very promising. KDE4.9.3 desktop in a 183 MB iso.
No trace of akonadi/nepomuk/strigi..etc. The dev has trimmed ALL the fat from KDE.
This is the first release from any distro to do this! Wonder why?
I have compiled gentoo kde4 without the fat, but Slax-7 is the ONLY public release, I know of, to do so.
It is a shame slax is not a conventional install. It is 'live' only, using modules to add apps. While it can be copied to a hard drive and booted, still not quite the same.
The release is not "usable in production" yet, as most of the modules still have to be compiled.
When slax 7 final release occurs, I expect it will draw a bit of attention.
46 • @ 45 - KDE fat (by Ralph on 2012-11-11 01:22:16 GMT from Canada)
I was under the impression that Kubuntu had a "kde-light" option that doesn't install Nepomuk, etc.?
47 • @ 46: kde-light (by RollMeAway on 2012-11-11 06:26:53 GMT from United States)
I have done a web search for "kde-light" and "kde-lite" .
Best I could discover is a couple of distros that leave out most of kde's applications, and call it lite.
If you can provide any link to such a creature, I'm sure many would be interested.
48 • There is a serious Phoronix Corruption Bug (by Trizt on 2012-11-11 08:04:50 GMT from Sweden)
Please delete the "Ext4 Data Corruption Bug and Solution", just bad to spread misinformation, phoronix is just making scandal journalism, writing things without really understand what he is writing about and blow out pseudo news like it would be the end of the world.
49 • 48 Imperfection Notification - it's a healthy thing (by Fossilizing Dinosaur on 2012-11-11 17:49:03 GMT from United States)
A serious data corruption programming error was noted, which reminds us all that a compendium of efforts of many imperfect human beings, no matter how careful, can still be imperfect.
That's a healthy thing for us to be mindful of, and reminded of.
Even though the conditions for this one may be rare, it's very real, and frighteningly serious. Maybe not the end of the world, but certainly the end of an entire computer system, and data entrusted to it. Constant vigilance is a required discipline, however occasionally inconvenient.
Minimizing the problem and conning others into a false sense of security should be vigorously discouraged. Sounding a timely warning, alerting the community, should not.
50 • @ 47 - kde-minimal (by Ralph on 2012-11-11 18:22:46 GMT from Canada)
I guess the package I had in mind was kde-minimal available for ubuntu Lucid (2 years ago), at least according to the following link, where a lot of people were calling it (erroneously, I think) 'kubuntu-minimal':
I did not read the discussion very carefully, so I don't know if Nepomuk, etc. came with this metapackage. In any case, it appears you have to start out with the ubuntu minimal-CD from the CLI and install the metapackage after the base system is installed. People in the discussion are saying that even thougth kde-minimal is no more, the same or similar result can be achieved by installing either kde-desktop with the no-recommends option or kde-plasma-desktop. I doubt that the former option would avoid installing Akonadi (isn't Akonadi a dependency of Kmail?), but the later I suspect would.
In any case, what Kubuntu needs is a minimal install option right on the regular DVD. It is the case (last time I checked) that by toggling the F6 options button at the beginning of the install you have the possibility of not installing the non-free and restricted repos, and this would be the logical place for activating the minimal install.
51 • #44 - date & time in openSUSE (by Caitlyn Martin on 2012-11-11 19:04:01 GMT from United States)
#44- Mac: the date command (at the command line) works exactly the same way in openSUSE as it does in any other distro. Setting the time to CST via Yast also works for me. Both do require the root password or sudo privileges just like any other distro. I have no clue what you mean about the forums or the distro not liking root. Can you explain that?
You like Debian better. That's fine. I prefer openSUSE to Debian and i think 12.2 is a fine release. That's also fine. Normally I dismiss "it works for me" comments and I can't think of the last time I made one, but here you are claiming basic functionality that is not distro specific doesn't work. I find that very hard to believe.
Let me give you the benefit of the doubt. What steps do you take when it fails? Can you detail them?
Oh, and before someone dismisses me as a fan, please look at the last openSUSE review I wrote for DistroWatch a couple of years back. I took a lot of heat for a truly negative review. That was then. This is now. 12.2 is the best openSUSE release I've seen in quite some time. It just works and works well on my hardware.
52 • @51 (by Mac on 2012-11-12 00:27:48 GMT from United States)
The short here is in the chair and thanks for the answer. I have overcome most of my problems with suse and working on the rest. It is a pleasure to see your comments here. Have been in debain so long that all the crossover doesn't seem to work. O I have reinstalled suse on primary partition and getting along better. But at my age who knows LOL.
Have fun Mack
53 • @51 (by Mac on 2012-11-12 01:27:06 GMT from United States)
After reinstall to primary partition doesn't behave the same, must have done something different. Makes my earlier comment seam a little out of line. Will be the first to admit that was a boob boob. But just found that out today or would have ask for a delete.
Have fun Mack
Number of Comments: 53
Display mode: DWW Only • Comments Only • Both DWW and Comments
|• Issue 828 (2019-08-19): AcademiX 2.2, concerns with non-free firmware, UBports working on Unity8, Fedora unveils new EPEL channel, FreeBSD phasing out GCC|
|• Issue 827 (2019-08-12): Q4OS, finding files on the disk, Ubuntu works on ZFS, Haiku improves performance, OSDisc shutting down|
|• Issue 826 (2019-08-05): Quick looks at Resilient, PrimeOS, and BlueLight, flagship distros for desktops,Manjaro introduces new package manager|
|• Issue 825 (2019-07-29): Endless OS 3.6, UBports 16.04, gNewSense maintainer stepping down, Fedora developrs discuss optimizations, Project Trident launches stable branch|
|• Issue 824 (2019-07-22): Hexagon OS 1.0, Mageia publishes updated media, Fedora unveils Fedora CoreOS, managing disk usage with quotas|
|• Issue 823 (2019-07-15): Debian 10, finding 32-bit packages on a 64-bit system, Will Cooke discusses Ubuntu's desktop, IBM finalizes purchase of Red Hat|
|• Issue 822 (2019-07-08): Mageia 7, running development branches of distros, Mint team considers Snap, UBports to address Google account access|
|• Issue 821 (2019-07-01): OpenMandriva 4.0, Ubuntu's plan for 32-bit packages, Fedora Workstation improvements, DragonFly BSD's smaller kernel memory|
|• Issue 820 (2019-06-24): Clear Linux and Guix System 1.0.1, running Android applications using Anbox, Zorin partners with Star Labs, Red Hat explains networking bug, Ubuntu considers no longer updating 32-bit packages|
|• Issue 819 (2019-06-17): OS108 and Venom, renaming multiple files, checking live USB integrity, working with Fedora's Modularity, Ubuntu replacing Chromium package with snap|
|• Issue 818 (2019-06-10): openSUSE 15.1, improving boot times, FreeBSD's status report, DragonFly BSD reduces install media size|
|• Issue 817 (2019-06-03): Manjaro 18.0.4, Ubuntu Security Podcast, new Linux laptops from Dell and System76, Entroware Apollo|
|• Issue 816 (2019-05-27): Red Hat Enterprise Linux 8.0, creating firewall rules, Antergos shuts down, Matthew Miller answers questions about Fedora|
|• Issue 815 (2019-05-20): Sabayon 19.03, Clear Linux's developer features, Red Hat explains MDS flaws, an overview of mobile distro options|
|• Issue 814 (2019-05-13): Fedora 30, distributions publish Firefox fixes, CentOS publishes roadmap to 8.0, Debian plans to use Wayland by default|
|• Issue 813 (2019-05-06): ROSA R11, MX seeks help with systemd-shim, FreeBSD tests unified package management, interview with Gael Duval|
|• Issue 812 (2019-04-29): Ubuntu MATE 19.04, setting up a SOCKS web proxy, Scientific Linux discontinued, Red Hat takes over Java LTS support|
|• Issue 811 (2019-04-22): Alpine 3.9.2, rsync examples, Ubuntu working on ZFS support, Debian elects new Project Leader, Obarun releases S6 tools|
|• Issue 810 (2019-04-15): SolydXK 201902, Bedrock Linux 0.7.2, Fedora phasing out Python 2, NetBSD gets virtual machine monitor|
|• Issue 809 (2019-04-08): PCLinuxOS 2019.02, installing Falkon and problems with portable packages, Mint offers daily build previews, Ubuntu speeds up Snap packages|
|• Issue 808 (2019-04-01): Solus 4.0, security benefits and drawbacks to using a live distro, Gentoo gets GNOME ports working without systemd, Redox OS update|
|• Issue 807 (2019-03-25): Pardus 17.5, finding out which user changed a file, new Budgie features, a tool for browsing FreeBSD's sysctl values|
|• Issue 806 (2019-03-18): Kubuntu vs KDE neon, Nitrux's znx, notes on Debian's election, SUSE becomes an independent entity|
|• Issue 805 (2019-03-11): EasyOS 1.0, managing background services, Devuan team debates machine ID file, Ubuntu Studio works to remain an Ubuntu Community Edition|
|• Issue 804 (2019-03-04): Condres OS 19.02, securely erasing hard drives, new UBports devices coming in 2019, Devuan to host first conference|
|• Issue 803 (2019-02-25): Septor 2019, preventing windows from stealing focus, NetBSD and Nitrux experiment with virtual machines, pfSense upgrading to FreeBSD 12 base|
|• Issue 802 (2019-02-18): Slontoo 18.07.1, NetBSD tests newer compiler, Fedora packaging Deepin desktop, changes in Ubuntu Studio|
|• Issue 801 (2019-02-11): Project Trident 18.12, the meaning of status symbols in top, FreeBSD Foundation lists ongoing projects, Plasma Mobile team answers questions|
|• Issue 800 (2019-02-04): FreeNAS 11.2, using Ubuntu Studio software as an add-on, Nitrux developing znx, matching operating systems to file systems|
|• Issue 799 (2019-01-28): KaOS 2018.12, Linux Basics For Hackers, Debian 10 enters freeze, Ubuntu publishes new version for IoT devices|
|• Issue 798 (2019-01-21): Sculpt OS 18.09, picking a location for swap space, Solus team plans ahead, Fedora trying to get a better user count|
|• Issue 797 (2019-01-14): Reborn OS 2018.11.28, TinyPaw-Linux 1.3, dealing with processes which make the desktop unresponsive, Debian testing Secure Boot support|
|• Issue 796 (2019-01-07): FreeBSD 12.0, Peppermint releases ISO update, picking the best distro of 2018, roundtable interview with Debian, Fedora and elementary developers|
|• Issue 795 (2018-12-24): Running a Pinebook, interview with Bedrock founder, Alpine being ported to RISC-V, Librem 5 dev-kits shipped|
|• Issue 794 (2018-12-17): Void 20181111, avoiding software bloat, improvements to HAMMER2, getting application overview in GNOME Shell|
|• Issue 793 (2018-12-10): openSUSE Tumbleweed, finding non-free packages, Debian migrates to usrmerge, Hyperbola gets FSF approval|
|• Issue 792 (2018-1203): GhostBSD 18.10, when to use swap space, DragonFly BSD's wireless support, Fedora planning to pause development schedule|
|• Issue 791 (2018-11-26): Haiku R1 Beta1, default passwords on live media, Slax and Kodachi update their media, dual booting DragonFly BSD on EFI|
|• Issue 790 (2018-11-19): NetBSD 8.0, Bash tips and short-cuts, Fedora's networking benchmarked with FreeBSD, Ubuntu 18.04 to get ten years of support|
|• Issue 789 (2018-11-12): Fedora 29 Workstation and Silverblue, Haiku recovering from server outage, Fedora turns 15, Debian publishes updated media|
|• Issue 788 (2018-11-05): Clu Linux Live 6.0, examining RAM consumpion, finding support for older CPUs, more Steam support for running Windows games on Linux, update from Solus team|
|• Issue 787 (2018-10-29): Lubuntu 18.10, limiting application access to specific users, Haiku hardware compatibility list, IBM purchasing Red Hat|
|• Issue 786 (2018-10-22): elementary OS 5.0, why init keeps running, DragonFly BSD enables virtual machine memory resizing, KDE neon plans to drop older base|
|• Issue 785 (2018-10-15): Reborn OS 2018.09, Nitrux 1.0.15, swapping hard drives between computers, feren OS tries KDE spin, power savings coming to Linux|
|• Issue 784 (2018-10-08): Hamara 2.1, improving manual pages, UBports gets VoIP app, Fedora testing power saving feature|
|• Issue 783 (2018-10-01): Quirky 8.6, setting up dual booting with Ubuntu and FreeBSD, Lubuntu switching to LXQt, Mint works on performance improvements|
|• Issue 782 (2018-09-24): Bodhi Linux 5.0.0, Elive 3.0.0, Solus publishes ISO refresh, UBports invites feedback, Linux Torvalds plans temporary vacation|
|• Issue 781 (2018-09-17): Linux Mint 3 "Debian Edition", file systems for SSDs, MX makes installing Flatpaks easier, Arch team answers questions, Mageia reaches EOL|
|• Issue 780 (2018-09-10): Netrunner 2018.08 Rolling, Fedora improves language support, how to customize Kali Linux, finding the right video drivers|
|• Issue 779 (2018-09-03): Redcore 1806, keeping ISO downloads safe from tampering, Lubuntu makes Calamares more flexible, Ubuntu improves GNOME performance|
|• Issue 778 (2018-08-27): GuixSD 0.15.0, ReactOS 0.4.9, Steam supports Windows games on Linux, Haiku plans for beta, merging disk partitions|
|• Issue 777 (2018-08-20): YunoHost 22.214.171.124, limiting process resource usage, converting file systems on Fedora, Debian turns 25, Lubuntu migrating to Wayland|
|• Full list of all issues|
Star Labs - Laptops built for Linux.
View our range including the Star Lite, Star LabTop and more. Available with a choice of Ubuntu, Linux Mint or Zorin OS pre-installed with many more distributions supported. Visit Star Labs for information, to buy and get support.
|Random Distribution |
Matriux was a Debian-based security distribution designed for penetration testing and forensic investigations. Although it was primarily designed for security enthusiasts and professionals, it can also be used by any Linux user as a desktop system for day-to-day computing. Besides standard Debian software, Matriux also ships with an optimised GNOME desktop interface, over 300 open-source tools for penetration testing, and a custom-built Linux kernel.