Markus Egg said the following on 02/06/2013 11:00 AM:

...

Von: Anton Aylward

...

*THAT* was your mistake. The 'defaults' are 'general purpose'. 'News' has specific needs.

I assumed that this was the mistake. I remembered that there were issues mentioned in the documentation of SuSE 10.3 and I thought that SuSE 12.1 takes care of that, but it did not. :-(

Well there is /etc/mke2fs.conf which "takes care of that" :-(

[...]

...

*MY* "solution' is not to use an ext file system. I use ReiserFS which is dynamic about its allocation. There are no fixed numbers of inodes. This is how things *should* be, in MNSHO. You should not be tied down to having to get your initialization parameters right. This is also why I use LVM - I don't want to have to get the size of the partition and FS 'right' to start with!

I thought of the following solution: 1) save fstab somewhere

Not actually needed. What you *DO* need is to be in single user/maintenance mode. That is much more important! All you're going to do with the fstab is ADD one entry for the new file system. That's all. I discuss this below.

2) rm the usenet part in /var/spool/news

Check

3) tar the rest of /var somewhere else (e.g. some external USB-disk)

Check. This is where my decision to use 5G partitions comes into play. Everything is made up out of 5G partitions that get backed up onto DVD. Well, not *everything*. There's a lot you don't need to back up if you're doing a reinstall :-)

4) use gparted to reboot into the gparted system and then reformat /var with reiserfs.

You can. This is where my decision to use LVM comes into play :-) I can shrink the file system - yes you can do that with ext4 - then shrink the partition, freeing up space for another partition. I suppose you could do the same with /var on a fixed size partition, using parted. Hmm Gparted requires multi-user mode. You don't want to be doing this in multi user mode! Why not? First, multi-user mode isn't going to be happy with /var unmounted! There are other processes that will want to access it. Again, this is why I favour ReiserFS and LVM. I can shrink or expand a ReiserFS without having to unmount it and I can alter size of a LVM partition to match the new reduced size of the FS without having to unmount the FS or go to single user mode. Similar for growing a partition & FS. Similar for adding a new partition and FS. No need to reboot to re-read the partition table either. Your situation is not that of a LVM. It is possible to resize the /var file system in single user mode and then shrink the partition, but its a *s*c*a*r*y* thing to try so the backup/restore is your best way. *IF* You are simply converting /var to a resierFS - which is the simplest thing here - then there's no need to resize the partitions, just backup, do the mkfs.resiserfs, then edit /etc/fstab to change the FS type of /var, mount and restore then go multi-user. *HOWEVER* is you are going to resize the partition for /var and create a new partition for /var/spool/news then you'll have to think about sizes - which while different than thinking about inode/data ratios is still a buqqer and why I chose to go with ReiserFS and LVM.

5) play back the /var content onto the new reiser-var

Not yet.

6) mount /etc with the gparted system

No. What you want is to create a new entry in /etc/fstab for the reiserFS with news. It will look something like this /dev/sda6 /var/spool/news reiserfs defaults 1 2 I can't tell what partition so I'm guessing sda6, you'll have to figure that yourself. ReiserFS can be mounted with options that will make it a bit faster, check the mount man page for these. Of course there's the issue of one partition or two.

7) But what do I have to enter into the /etc/fstab manually afterwards? Put the fstab onto /etc

See above.

8) if 7) is done reboot to SuSE 12.1

Its the resizing of the partition and the creation of the new partition that necessitates this.

9) transfer the /var/spool/news onto the new /var under SuSE 12.1

Is that feasible? I already did some 200GB -> 500GB enlargements with gparted and SuSE 10.3 . It was tricky but it worked.

Maybe I'm confused here. Some of what you wrote indicates that you're creating a new news partition, some of it that you're not and are just converting /var to a ResiserFS. Or are you creating a new partition with a ReiserFS for /var/spool/news *AND* converting /var itself to a ResiserFS? In all of the above I'd strongly suggest reading the man pages and googling around to clarify anything you're not clear about BEFOREHAND. Now you know why I'm experimenting with a single FS using BtrFS and nothing else: no /boot; no worries about merging / and /usr; no worries about inode/data ratios; no worries about barriers. -- If you can't describe what you are doing as a process, you don't know what you're doing. -- W. Edwards Deming -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org

Markus Egg said the following on 02/06/2013 12:19 PM:

...

Datum: Wed, 06 Feb 2013 12:08:39 -0500 Von: Anton Aylward An: opensuse@opensuse.org Betreff: Re: [opensuse] Problem with inodes on SuSE 12.1 and /var ?

[...]

...

Maybe I'm confused here. Some of what you wrote indicates that you're creating a new news partition, some of it that you're not and are just converting /var to a ResiserFS.

Or are you creating a new partition with a ReiserFS for /var/spool/news *AND* converting /var itself to a ResiserFS?

I am only converting complete /var to ReiserFS and keep the same size for /var.

The things become very very simple. FORGET (g)parted. Its an irrelevance now. Its only relevant if you are altering partition sizes or creating a new partition. 1. Go into single user mod. 2. Delete below /var/spool/news 3. Back up /var 4. unmount /var 5. Run mkfs.resiserfs on the /var/partition 6. Change the /var entry in /etc/fstab 7. Remount /var 8. Restore /var 9. Go to multi user mode 10 Start news

I do not want to make some extra partition for /var/spool/news if that is possible (btw. that worked for SuSE 10.3 and ext3). So going back to ext3 for /var would be another options as I assume that SuSE 12.1 handles /var the same way as SuSE 10.3 with ext3(correct?).

I doubt it very much. You'd have the same problem about inodes vs data that you had with ext4.

Other possible problem: does gpart have the same options for ReiserFS as SuSE 12.1? Otherwise I will get into funny trouble...

FORGET gpart/parted. Its only relevant if you are altering partition size or creating a new partition.

...

In all of the above I'd strongly suggest reading the man pages and googling around to clarify anything you're not clear about BEFOREHAND.

Now you know why I'm experimenting with a single FS using BtrFS and nothing else: no /boot; no worries about merging / and /usr; no worries about inode/data ratios; no worries about barriers.

Yes but btrfs is still very experimental afaics.

From my POV ext4 is still experimental. So what? I'm an engineer; it works for me. I'm kicking it around and not having any problems with it and getting to understand it better than I did last year.

-- Information is the currency of democracy. --Thomas Jefferson -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org

Markus Egg said the following on 02/10/2013 07:04 AM:

Am 06/02/13 18:44, schrieb Anton Aylward:

[...]

...

FORGET (g)parted. Its an irrelevance now. Its only relevant if you are altering partition sizes or creating a new partition.

1. Go into single user mod. 2. Delete below /var/spool/news 3. Back up /var 4. unmount /var 5. Run mkfs.resiserfs on the /var/partition 6. Change the /var entry in /etc/fstab 7. Remount /var 8. Restore /var 9. Go to multi user mode 10 Start news [...]

Thanks very much, I will try this. Do I just have to change the type of the /var entry in /etc/fstab or additonally something else?

If, as you say, all you are doing is changing the FS type of the partition, then all you have to do is change the FStype entry for that partition in the fstab file before remounting[1]. If you are adding a partition, as I first thought, mistakenly it seems, then you would need to add an entry. But if you're not then you don't. Yes it really is that simple and straight forward. The important thing is to do it in single user mode - all those background processes disabled, otherwise you can't unmount /var. [1] If you don't the system will try mounting with the wrong FS type and that will fail and you'll end up running without /var mounted and that will cause a cascade of errors. -- Opportunities multiply as they are seized. --Sun Tzu -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org

Anton Aylward wrote:

...

...

Or are you creating a new partition with a ReiserFS for /var/spool/news *AND* converting /var itself to a ResiserFS? I am only converting complete /var to ReiserFS and keep the same size for /var. I doubt it very much. You'd have the same problem about inodes vs data that you had with ext4.

Give the above, why not switch to a file system that allows you change the percentage space given to inodes, 'after the fact'? I don't know which file systems support this besides 'xfs' (using xfs_growfs -m). Seems like that might be worth some minor consideration.

From my POV ext4 is still experimental. So what? I'm an engineer; it works for me. I'm kicking it around and not having any problems with it and getting to understand it better than I did last year.

And from this perspective, there are not any file systems on linux with more maturity behind them than xfs. I think it's 20 years old this year, yet it's beginning feature set included more features than what even many current file systems support by default or at all, including ACL's, XATTR's, real-time IO, high file I/O rates, generally low fragmentation, but one of the few to come with a file "defrag'ger" for fixing degenerate cases (literally -- that's why it was written -- a customer came to them and wanted to know why he wasn't getting the perf he should have -- his app generated lots of file interleaving -- not good for single-file I/O... According to the xfs website, openSuSE is one of the distros that works with the xfs team to make sure it works on openSuSE systems. Caveat: not for use on systems where sudden power loss and crashes are frequent! Works excellently if you have a UPS or similar (a laptop was great, as any power outage, it automatically went to battery). XFS keeps much stuff in buffers as long as possible so it can be written to disk in as few as disk writes as possible. The default on desktops is about 5 seconds (in computer time, that's a long time), but in laptop mode, it would flush about every 30 so worked well there, as well... Anyway -- there may be other FS that support reallocating inode space, they might be more convenient than running into the same problem again later. -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org

Anton Aylward wrote:

Let me get this right.

As with the extN file systems, you can grow XFS. But you still get the situation the OP met, the 'out of space' message even though 'df' says there is space .... and you scratch your head and then finally go 'Ahh!' and run 'df -i' and see that Iuse% is 100. Then you "grow" -- change - the percentage of the FS given to inodes.

Ah, but the man page says In order to grow a filesystem, it is necessary to provide added space for it to occupy.

Can you alter the inode %age "in place"?

Yes. But the file system has to be mounted... FWIW -- I didn't know, so I had to try it. # xfs_growfs /dev/mapper/HnS-Sys -m 6 meta-data=/dev/mapper/HnS-Sys isize=256 agcount=16, agsize=1572848 blks = sectsz=512 attr=2 data = bsize=4096 blocks=25165568, imaxpct=5 = sunit=16 swidth=192 blks naming =version 2 bsize=4096 ascii-ci=0 log =internal bsize=4096 blocks=32768, version=2 = sectsz=512 sunit=16 blks, lazy-count=1 realtime =none extsz=4096 blocks=0, rtextents=0 data blocks changed from 25165568 to 25165824 inode max percent changed from 5 to 6

As to the 'oldest', well the ReiserFS was integrated into Linux before XFS was, and BTree file systems have a long history. Go google. And please don't try to tell me that the idea was based upon Microsoft's NTFS. IBM's JFS and the VERITAS File System both predate XFS. To be fair, there was a hefty burst of development in file systems for UNIX variants in the early 1990s.

XFS dates back to the early 90's. It was ported to linux, but not created new. -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org

Peter Maffter said the following on 02/25/2013 05:43 AM:

...

Von: Anton Aylward An: opensuse@opensuse.org

...

Gesendet: 18:44 Mittwoch, 6.Februar 2013 Betreff: Re: [opensuse] Problem with inodes on SuSE 12.1 and /var ?

...

1. Go into single user mod. 2. Delete below /var/spool/news 3. Back up /var 4. unmount /var 5. Run mkfs.resiserfs on the /var/partition 6. Change the /var entry in /etc/fstab 7. Remount /var 8. Restore /var 9. Go to multi user mode 10 Start news

Does this also work on a SSD with reiserfs and /var/spool/news ?

The LOGIC of the above procedure will work for any file system, not just reiserFS. Will it work for var/spool/news ? It will work for any sub-tree you want. I've done it for /usr/lib/ruby and /usr/lib/perl5 and /usr/share among others. Where a partition gets mounted is just an entry in /etc/fstab and has nothing to do, other than the values in the fields, with the where or FS or source. RTFM.

Any precautionary measure when using reiserfs on a SSD?

I've not used SSD. On this point Id advise 'Go Google". Its one of the reasons I'm experimenting with BtrFS. The bad thing about BtrFS is that there is a lot of old and out of date articles on the Net-of-a-Million-Lies saying its experimental and unstable. Big Deal. At one time so was ext4 .... and ext3 for that matter if you go back far enough! I'm finding BtrFS quite stable. The problems I'm having with it are problems with my limited understanding. -- Telling the future by looking at the past assumes that conditions remain constant. This is like driving a car by looking in the rear view mirror. - Herb Brody -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org

On Tue, 16 Apr 2013 05:03:18 +0200 (CEST) "Carlos E. R." wrote:

it probably writes often to the same disk area

Controller decides where to write, in order to keep disk alive as long as possible. Also, having bigger disk then minimum needed is advantage as controller can spread wear more evenly. The best way to kill SSD fast is to fill it up with static content and leave little space to dynamic, like cache and temporary files. Part of the drive that is changing all the time will reach end of life fast, while the rest is still in a good shape. -- Regards, Rajko. -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 On Tuesday, 2013-04-16 at 07:46 -0400, Greg Freemyer wrote:

"Carlos E. R." <> wrote:

That was true in 2005. In 2013 pretty much all SSDs and even thumb drives have wear leveling logic.

But not enough. According to https://wiki.linaro.org/WorkingGroups/KernelArchived/Projects/FlashCardSurvey?action=show&redirect=WorkingGroups%2FKernel%2FProjects%2FFlashCardSurvey "Flash memory card design", most portable flash media is optimized for FAT in several ways. For example, the start of the disk, the region where the FAT would be, is optimized for the expected use of a FAT table. There is a new filesystem type that is optimized for use in cheap flash media: f2fs, introduced by Samsung developers in October. http://www.h-online.com/open/features/Kernel-Log-Coming-in-3-8-Part-1-Filesy... Kernel Log - Coming in 3.8 (Part 1) Filesystems and storage http://lwn.net/Articles/518988/ An f2fs teardown - -- Cheers, Carlos E. R. (from 12.1 x86_64 "Asparagus" at Telcontar) -----BEGIN PGP SIGNATURE----- Version: GnuPG v2.0.18 (GNU/Linux) iEYEARECAAYFAlFteKYACgkQtTMYHG2NR9UALQCgk6Q4Q35pXAWkYG007ecwiuUZ i30AnAs91FbT0glmmTm36eK5CClkc2MP =wej7 -----END PGP SIGNATURE----- -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org

Carlos E. R. said the following on 02/10/2013 09:56 AM:

True. But I heard it has an interesting feature: it can do compression of files, and news are text files. I will eventually switch a partition to it and try.

Compression? It seems so; I've just installed a 12.2 on a 20G drive with BtrFS alone, no separate bot partition just one long BtrFS partition, and there's enough space left over to convince me its compressed enough to install and run on a dinky little 8G drive and still have something left over. And that's with LibreOffice, all the KDE stuff ... So something like a Raspberry Pi with two USB ports ... one for the system one to have /home. No, wait, a 32G usb stick -- overkill! -- For ages, a deadly conflict has been waged between a few brave men and women of thought and genius upon the one side, and the great ignorant religious mass on the other. This is the war between Science and Faith. The few have appealed to reason, to honor, to law, to freedom, to the known, and to happiness here in this world. The many have appealed to prejudice, to fear, to miracle, to slavery, to the unknown, and to misery hereafter. The few have said "Think" The many have said "Believe!" --Robert Ingersoll (Gods) -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org