| View previous topic :: View next topic |
| Author |
Message |
notHerbert
Advocate
Joined: 11 Mar 2008
Posts: 2228
Location: 45N 73W
|
 Posted: Sat May 03, 2008 2:10 pm Post subject: |
 |
|
I use JFS on my laptop. Real nice and super fast.
My /usr/portage is ext2 - 3G - with eclean-dist -d it is 50% full
My /var/tmp is ext2 - 5G
My /boot is ext2
All the rest is JFS.  |
|
| Back to top |
|
 |
nixnut
Bodhisattva
Joined: 09 Apr 2004
Posts: 10974
Location: the dutch mountains
|
| Posted: Sat May 03, 2008 2:21 pm Post subject: |
|
|
merged above two posts here
_________________
Please add [solved] to the initial post's subject line if you feel your problem is resolved. Help answer the unanswered
talk is cheap. supply exceeds demand |
|
| Back to top |
|
|
Beju
Apprentice
Joined: 30 Mar 2006
Posts: 171
|
| Posted: Fri May 23, 2008 5:05 pm Post subject: Which FS do you recommend? |
|
|
Hello,
I would like to know your adivces on which filesystem should I choose for my new disk. The disk is meant to be used mainly as a storage, with 200MB and larger files on it, so I wouldn't want to lose additional disk space needlessly. Also, it doesn't have to be very fast, but the FS should be stable, with low chance to crash or data loss. |
|
| Back to top |
|
|
leealee
n00b
Joined: 16 Oct 2007
Posts: 22
Location: England
|
| Posted: Fri May 23, 2008 5:33 pm Post subject: |
|
|
| ReiserFS I have always found to be fast and stable. |
|
| Back to top |
|
|
bunder
Bodhisattva
Joined: 10 Apr 2004
Posts: 5934
|
| Posted: Fri May 23, 2008 5:34 pm Post subject: |
|
|
merged two posts.
_________________
| Neddyseagoon wrote: | | The problem with leaving is that you can only do it once and it reduces your influence. |
banned from #gentoo since sept 2017 |
|
| Back to top |
|
|
nonhuman
Apprentice
Joined: 30 Sep 2002
Posts: 236
Location: Washington DC
|
| Posted: Mon May 26, 2008 5:11 pm Post subject: |
|
|
Ok, I'm planning out a Gentoo install for my new ThinkPad that should be arriving in the next couple days. I've never really been a big fan of ext*, and I'm not sure I want to use reiser* with it's uncertain future. I'm very interested in btrfs, however, though I'm not sure I really want to use it on anything critical just yet. So for the time being I'm thinking I'll use a mixed JFS/btrfs system. My understanding of the way btrfs works, is that I can pretty much just create a single partition and then use subvolumes for the individual mountpoints the way one would use a ZFS zpool, and that way I don't need to worry about allocated space specifically to the individual mountpoints, is that right?
I'm thinking /, /boot, and /home will be JFS, while /usr/portage, /var/tmp, and /usr/src will be btrfs. If I'm understanding things correctly that should just be four partitions (plus swap, obviously).
Does that make sense? Are there any other places that it would make sense to use btrfs? I'm kind of considering using it for /home just because most of my files will be code that are safely backed up in a SVN repository on my server.
_________________
"Ambition is a poor excuse for not having sense enough to be lazy." -Edgar Bergen |
|
| Back to top |
|
|
easy target
Tux's lil' helper
Joined: 09 Jan 2008
Posts: 134
|
| Posted: Mon May 26, 2008 6:38 pm Post subject: |
|
|
This maybe little offtopic, but you can take a look at lvm2, that should let you manage your partitions much more easily than std partitions. |
|
| Back to top |
|
|
energyman76b
Advocate
Joined: 26 Mar 2003
Posts: 2048
Location: Germany
|
| Posted: Wed May 28, 2008 8:38 am Post subject: |
|
|
JFS is slow and nothing happening there - close to unsupported.
If you care about your data, choose reiserfs. Don't choose ext3. Don't choose lvm. XFS is ok too, but there are too many fuck ups for my taste (every three month something happens that screws XFS users ...).
_________________
Study finds stunning lack of racial, gender, and economic diversity among middle-class white males
I identify as a dirty penismensch. |
|
| Back to top |
|
|
nonhuman
Apprentice
Joined: 30 Sep 2002
Posts: 236
Location: Washington DC
|
| Posted: Wed May 28, 2008 4:05 pm Post subject: |
|
|
| energyman76b wrote: | JFS is slow and nothing happening there - close to unsupported.
If you care about your data, choose reiserfs. Don't choose ext3. Don't choose lvm. XFS is ok too, but there are too many fuck ups for my taste (every three month something happens that screws XFS users ...). |
Really? I've never heard anyone say anything about JFS other than that it's really fast. I've definitely heard enough bad things about XFS that I'm not really inclined to try it out.
I suppose I could put all my filesystems on reiser, but that hardly seems like it's going to be better than JFS in terms of support and ongoing development in the future, and from what I've read it sounds less reliable.
Why would you recommend against LVM?
_________________
"Ambition is a poor excuse for not having sense enough to be lazy." -Edgar Bergen |
|
| Back to top |
|
|
energyman76b
Advocate
Joined: 26 Mar 2003
Posts: 2048
Location: Germany
|
| Posted: Wed May 28, 2008 8:59 pm Post subject: |
|
|
reiserfs has almost nil bug reports on lkml or reiser-ml. It is stable. Almost all horror stories are from early 2.4 times when the constant vm changes broke reiserfs - and the guys who broke it could not be assed to repair the damage they did.
If you look at here http://bulk.fefe.de/lk2006/bench.html jfs is not really fast. Add to that mix the lack of users and a 'maintanance' mode that makes reiserfs looks like a patch-addicted feature whore. One patch in years. The lack of problem reports is mostly based on the fact that almost nobody uses it.
reiserfs&xfs use barriers by default. You can turn them on for ext3 and get a 30% performance decrease. Barriers are important with todays huge hard disk caches - and disabling the cache is even worse then enabling barriers. If you care about your data, you want barriers. But LVM does not support barriers. Even if the fs ontop of it uses them, they don't 'reach down' - and that is a BAD THING™.
http://marc.info/?l=linux-fsdevel&m=121097127009793&w=2
http://marc.info/?l=linux-fsdevel&m=121097545716390&w=2
But XFS is at the moment (or since its introducion to the linux kernel) where reiserfs was in early 2.4 - it is broken on a monthly basis. Badly broken.
_________________
Study finds stunning lack of racial, gender, and economic diversity among middle-class white males
I identify as a dirty penismensch. |
|
| Back to top |
|
|
Enlight
Advocate
Joined: 28 Oct 2004
Posts: 3519
Location: Alsace (France)
|
| Posted: Mon Jun 02, 2008 10:01 pm Post subject: |
|
|
| arth1 wrote: | | Akkara wrote: | | Quote: | | How safe is XFS if the box will go down? |
A few years back when I researched this, I had read many reports of data loss with xfs on power outage, and even a subtle kernel bug around the -- I think -- with one of the 2.6.17's.
My own experience, when intentionally killing power to test what happens, is that the last bunch of files written will have 0's in them. The files will be there, but the data will be nulled. |
This is a feature, and not a flaw. Other file systems might leave the files as is, and they may or may not work afterwards, but XFS will
|
this part is right, the rest is just utterly false.
| Quote: |
zero them out because they might contain highly confidential data.
Scenario: CEO discovers a paragraph in a text file containing trade secrets, and deletes it, and hits save. Due to the nature of journalling file systems, the change goes to the log, and not directly to the file.
|
XFS only logs metadata (it's own metadata, not what you can see through the stat command) not data themselves. AFAIK only ext3 does this (partially because of log size limitation) when asked to (reiser4 atomic transactions are different from data logging).
| Quote: |
However, the system crashes. What should happen when the log replays? Other file systems will discard the incomplete journal entry, and leave the file as it was. XFS will zero out the file |
it doesn't fill the file with zero it deletes the node and returns data location to the free spool, xfs never deletes data, it deletes (as most filesystems) nodes entries and leave orphan data and metadata in the free spool where they might (or not) be overwriten.
There is no willing to protect secret datas, the goal mostly is to prevent the apparition of sparse files.
| Quote: |
in its entirety, not knowing what changes were to be made, only that it was changed. This is a plus from a security point of view, but for a "normal" user, it can be a PITA.
If using xfs without high availability, I strongly recommend learning how to use xfsdump and xfsrestore.
With HDs as cheap as they are, adding local storage for full and incremental backups should be within most people's means. |
In fact real user problem is not amputation of a file which was opened for modification, as it's application job to backup a file you're working on. The real problem is when it' comes to directory amputation. Directory data holds file names, so when unsure of directory datas states, xfs deletes the node pointing to the given data, and then the whole subtree becomes inconsistent and "has to" be deleted too. This could be changed by making changes in directory data atomic but it would totally break compatibility with actual version of xfs which would be unacceptable for commercial purpose |
|
| Back to top |
|
|
energyman76b
Advocate
Joined: 26 Mar 2003
Posts: 2048
Location: Germany
|
| Posted: Mon Jun 02, 2008 10:40 pm Post subject: |
|
|
| Enlight wrote: |
XFS only logs metadata (it's own metadata, not what you can see through the stat command) not data themselves. AFAIK only ext3 does this (partially because of log size limitation) when asked to (reiser4 atomic transactions are different from data logging).
|
wrong. reiserfs also 'logs' data.
At the end: if you care about your data, you can choose between reiserfs and reiserfs. xfs does not protect anything and likes to destroy stuff. It also breaks with every other kernel version. ext3 does not make sure that the stuff really ends on the platter.
_________________
Study finds stunning lack of racial, gender, and economic diversity among middle-class white males
I identify as a dirty penismensch. |
|
| Back to top |
|
|
Enlight
Advocate
Joined: 28 Oct 2004
Posts: 3519
Location: Alsace (France)
|
| Posted: Tue Jun 03, 2008 7:51 pm Post subject: |
|
|
| energyman76b wrote: | | Enlight wrote: |
XFS only logs metadata (it's own metadata, not what you can see through the stat command) not data themselves. AFAIK only ext3 does this (partially because of log size limitation) when asked to (reiser4 atomic transactions are different from data logging).
|
wrong. reiserfs also 'logs' data.
|
Thats' 99% wrong!
Reiserfs only stores formated blocks in it's journal, so by design it's a metadata journal. But some leaf nodes of the S-Tree may hold directly the file data in the case of files smaller than blocksize (i.e. 4kb). So when such a transaction is interupted before the flush is complete, the whole transaction can be replayed.
There seems to be another particular case when this is possible, it is the case of file spread on two data blocks where first is indirect item (+ datas) and second holds a tail (which mean that this scenario can't happen with "notail").
In fact, by design, reiserfs log mechanism is a metada only log, or if you rather, the reason why a block is put into the log, is only because it contains metadata. But as the (old) design of the tree packs keys other metadata and sometime data (see difference beetween B-Tree and B+tree) there are some particular cases where reiserfs can provide results similar to data logging.
This would be impossible if reiserfs was using B+Tree as xfs or reiser4 do.
| Quote: |
At the end: if you care about your data, you can choose between reiserfs and reiserfs. xfs does not protect anything and likes to destroy stuff. It also breaks with every other kernel version. ext3 does not make sure that the stuff really ends on the platter. |
 |
|
| Back to top |
|
|
energyman76b
Advocate
Joined: 26 Mar 2003
Posts: 2048
Location: Germany
|
| Posted: Tue Jun 03, 2008 8:18 pm Post subject: |
|
|
so? then explain the data=journal option for reiserfs.
And while you at it, you can also explain why reiserfs is safer because it does not ignore barriers like ext3.
EDIT: correction, ext3 does not ignore barriers. They are turned of by default. reiserfs and xfs have them turned on by default.
_________________
Study finds stunning lack of racial, gender, and economic diversity among middle-class white males
I identify as a dirty penismensch. |
|
| Back to top |
|
|
Enlight
Advocate
Joined: 28 Oct 2004
Posts: 3519
Location: Alsace (France)
|
| Posted: Tue Jun 03, 2008 9:52 pm Post subject: |
|
|
| energyman76b wrote: |
so? then explain the data=journal option for reiserfs.
And while you at it, you can also explain why reiserfs is safer because it does not ignore barriers like ext3.
EDIT: correction, ext3 does not ignore barriers. They are turned of by default. reiserfs and xfs have them turned on by default. |
Ok, just went through the sources, and it and saw this data=journal feature, I'm a bit astonished I never heard of it before. Where did you heard about it? |
|
| Back to top |
|
|
energyman76b
Advocate
Joined: 26 Mar 2003
Posts: 2048
Location: Germany
|
| Posted: Tue Jun 03, 2008 10:06 pm Post subject: |
|
|
| Enlight wrote: | | energyman76b wrote: |
so? then explain the data=journal option for reiserfs.
And while you at it, you can also explain why reiserfs is safer because it does not ignore barriers like ext3.
EDIT: correction, ext3 does not ignore barriers. They are turned of by default. reiserfs and xfs have them turned on by default. |
Ok, just went through the sources, and it and saw this data=journal feature, I'm a bit astonished I never heard of it before. Where did you heard about it? |
years ago on lkml, different linux sites ... it is an old feature. Chris Mason added it, if you want to google for the exact time.
http://www.linux-magazine.com/w3/issue/65/File_Systems.pdf
says that all three ext3 journal-options are also available for reiserfs. And super.c says the same 
_________________
Study finds stunning lack of racial, gender, and economic diversity among middle-class white males
I identify as a dirty penismensch. |
|
| Back to top |
|
|
deathcon1
Apprentice
Joined: 30 Aug 2007
Posts: 182
Location: Canada
|
| Posted: Tue Jun 24, 2008 5:50 pm Post subject: Best Filsystem |
|
|
| So I'm wondering what the best filesystem is for a laptop? Right now I'm running EXT2 "/boot" and JFS "/", but wondering what the best filesystem is for desktop/raid and obviously my laptop. Thoughts? |
|
| Back to top |
|
|
avieth
Veteran
Joined: 17 Sep 2004
Posts: 1945
Location: Canada
|
| Posted: Tue Jun 24, 2008 6:13 pm Post subject: |
|
|
Ext3 is the most stable solution for /. You might want to try reiserfs, it's good too.
Seriously though, you probably won't notice any difference in speed between filesystems on a desktop. You can't go wrong with ext3 or reiserfs. |
|
| Back to top |
|
|
nixnut
Bodhisattva
Joined: 09 Apr 2004
Posts: 10974
Location: the dutch mountains
|
| Posted: Tue Jun 24, 2008 6:24 pm Post subject: |
|
|
merged above two posts here.
_________________
Please add [solved] to the initial post's subject line if you feel your problem is resolved. Help answer the unanswered
talk is cheap. supply exceeds demand |
|
| Back to top |
|
|
deathcon1
Apprentice
Joined: 30 Aug 2007
Posts: 182
Location: Canada
|
| Posted: Tue Jun 24, 2008 9:10 pm Post subject: |
|
|
I've heard mixed things about reiser, mainly the closer it gets to being full the more fragmented it gets and it's nearly unrepairable at that point.
What are the pitfalls of JFS? |
|
| Back to top |
|
|
avieth
Veteran
Joined: 17 Sep 2004
Posts: 1945
Location: Canada
|
| Posted: Tue Jun 24, 2008 9:31 pm Post subject: |
|
|
| deathcon1 wrote: | I've heard mixed things about reiser, mainly the closer it gets to being full the more fragmented it gets and it's nearly unrepairable at that point.
What are the pitfalls of JFS? |
Check out this article. It's got some benchmark graphs.
| Quote: | | For those of you still reading, congrats! The conclusion is obvious by the "Total Time For All Benchmarks Test." The best journaling file system to choose based upon these results would be: JFS, ReiserFS or XFS depending on your needs and what types of files you are dealing with. I was quite surprised how slow ext3 was overall, as many distributions use this file system as their default file system. Overall, one should choose the best file system based upon the properties of the files they are dealing with for the best performance possible! |
They all have their pros and cons, but I still say you won't notice them in day-to-day tasks. |
|
| Back to top |
|
|
cold
n00b
Joined: 23 Feb 2003
Posts: 66
|
| Posted: Fri Jun 27, 2008 8:57 pm Post subject: |
|
|
In a few day's I get my new fileserver witch gives me a mind blowing 9TB disk space in a hardware raid5 (10X1TB)
The controller (a Dell PERC6E) will be capable of online raid extension and will go with a maximum of 45 disks well over any
16TB limit.
The challenge is that I want a stable and reliable filesystem witch is capable to grow with the array up to at least 30TB
The system I'm gonna use will run a 64bit 2.6.25 or 2.6.26 kernel (depends on when 2.6.26 is ready)
Is filled with files for 99.9% over 100MB and up to about 40GB
The array will use a GPT partition table as the partition will be way over 2TB
I have done some research and found 3 ext3, ext4, xfs, reiserfs
according to http://en.wikipedia.org/wiki/Comparison_of_file_systems#Limits
ext3 as a 32TB limit but the ext3 wiki says's 16TB witch just won't be enough, on the other hand it's proven tech has the ability to grow and performs reasonably well.
ext4 has a limit of 4EB according to wikipedia but it's still under development and i can't seem to find out if i can grow it.
and then there is XFS witch can do 8EB and growing partitions.
but I've never used it and have no idea about its stability and how robust it is although xfs should give the best performance with big files.
Reiserfs (3.6) will reiser 4 ever be final and stable with Reiser himself in jail?
no support for volumes over 16TB but does support growing a volume.
I will have some time to test (about 5 day's) but it would be nice to have that left for who knows what.
The files on this server are not mission critical but it is to much to backup and its really necessary to grow the filesystem reasonable save.
Any suggestions? |
|
| Back to top |
|
|
energyman76b
Advocate
Joined: 26 Mar 2003
Posts: 2048
Location: Germany
|
| Posted: Fri Jun 27, 2008 9:52 pm Post subject: |
|
|
| cold wrote: | I
Reiserfs (3.6) will reiser 4 ever be out with Reiser himself in jail?
no support for volumes over 16TB but does support growing a volume.
|
since extX does not do barriers by default and has a 30% penalty in performance - it might not be a good choice. if you use lvm, barriers are ignored anyway, so it might be back in the game if you use that.
reiser4 is 'out'. It is in mm - and for every kernel release there is a reiser4 patch. Compression is nice. But - it is still in development. You could ask Edward what still has to be done
As long as you are only using stable kernels - and wait some weeks between release of a kernel and switiching to it AND a good UPS, XFS should be fine. The main problem with XFS at the moment: every month there is a big, ugly bug reported with XFS on lkml. So if you use XFS, you should regularly check the list.
_________________
Study finds stunning lack of racial, gender, and economic diversity among middle-class white males
I identify as a dirty penismensch. |
|
| Back to top |
|
|
kernelOfTruth
Watchman
Joined: 20 Dec 2005
Posts: 6111
Location: Vienna, Austria; Germany; hello world :)
|
|
| Back to top |
|
|
cold
n00b
Joined: 23 Feb 2003
Posts: 66
|
| Posted: Fri Jun 27, 2008 10:05 pm Post subject: |
|
|
| energyman76b wrote: | | cold wrote: | I
Reiserfs (3.6) will reiser 4 ever be out with Reiser himself in jail?
no support for volumes over 16TB but does support growing a volume.
|
since extX does not do barriers by default and has a 30% penalty in performance - it might not be a good choice. if you use lvm, barriers are ignored anyway, so it might be back in the game if you use that.
reiser4 is 'out'. It is in mm - and for every kernel release there is a reiser4 patch. Compression is nice. But - it is still in development. You could ask Edward what still has to be done
As long as you are only using stable kernels - and wait some weeks between release of a kernel and switiching to it AND a good UPS, XFS should be fine. The main problem with XFS at the moment: every month there is a big, ugly bug reported with XFS on lkml. So if you use XFS, you should regularly check the list. |
A UPS is on it's way as we speak so that won't be the problem .
ill start checking the xfs bug report.... |
|
| Back to top |
|
|
|
Gentoo Chat
