On 03/11/12 13:19, Jason White wrote:

Terry Duell <tduell@iinet.net.au> wrote:

...

Are there issues with running Linux from an SSD, that one needs to take into account when first installing/partitioning/whatever?

It has been suggested here in the past that you should use the noatime option to ensure that not every access to a file causes blocks to be written to the ssd.

Given that the 'relatime' option is now the default (since kernel 2.6.30), it's probably not that much of a concern anymore. Yes, relatime will create slightly more writes than noatime would, but it will be considerably fewer than the old default option. (relatime only updates the access time if the previous access time was earlier than the current modify/change times; noatime causes programs like mutt to break) Going back to the original question ... I've had an SSD in a netbook for over two years now, Linux works fine on it. Cheers, Paul -- Paul Dwerryhouse | PGP Key ID: 0x6B91B584 http://weblog.leapster.org/

On 03/11/12 20:48, Jason White wrote:

Whether one would run Btrfs on a laptop at this stage in its development is a matter of judgment.

FWIW I've been running it on my work laptops for /home since January 2009 (pre-mainline kernel merge) and haven't had a problem (yet) with it. That does not mean it will be pain free for others, I don't use any features like snapshots, subvolumes, etc and it's done well by me so far. This page is worth reading before deciding if you want to try it out or not: https://btrfs.wiki.kernel.org/index.php/Gotchas cheers, Chris -- Chris Samuel : http://www.csamuel.org/ : Melbourne, VIC

On Sun, 4 Nov 2012, Andrew McGlashan <andrew.mcglashan@affinityvision.com.au> wrote:

I've subscribed to linux-btrfs-owner@vger.kernel.org mailing list, just to have an idea as to what is going on .... patches keep on coming and other issues crop up too often as well. I'm not ready to use BTRFS on anything yet myself. In time I do hope it turns out to be a better option than ZFS. I wish that there weren't any licensing issues with ZFS code living in the Linux kernel because I love ZFS on Solaris and I don't want to use ZFS via FUSE -- nor do I want to use BSD kernel (which is another possibility with Debian).

One major problem with ZFS is the way it manages memory. It's supposed to be possible to keep it's memory use down by limiting ARC size, but that doesn't seem to work as designed. My experience is that a zfsonlinux system with 4G of RAM and light load was giving kernel panics due to memory allocations failing. While it's relatively cheap to add heaps of RAM on new systems it's still an annoyance and this prevents the use of ZFS on small systems. So while I could probably upgrade my Thinkpad to 8G of RAM and ZFS but it's a lot easier to keep the current 5G and use BTRFS. It's possible that the FUSE option would improve things, presumably the user- space ZFS code run by FUSE would be pageable and therefore the memory allocation problems wouldn't be as bad. But FUSE has other issues and I'd prefer not to use it. -- My Main Blog http://etbe.coker.com.au/ My Documents Blog http://doc.coker.com.au/

On 4/11/2012 3:35 PM, Jason White wrote:

Andrew McGlashan <andrew.mcglashan@affinityvision.com.au> wrote:

...

I've subscribed to linux-btrfs-owner@vger.kernel.org mailing list, just to have an idea as to what is going on .... patches keep on coming and other issues crop up too often as well. I'm not ready to use BTRFS on anything yet myself.

Actually, this is the list I've subscribed: linux-btrfs@vger.kernel.org

Patches are to be expected as performance improvements and new features are added. It's the incidence of corruption not due to hardware issues that I would consider to be more indicative.

Okay, but I think it will be a while yet until I start to use BTRFS.... Kind Regards AndrewM

On 04/11/12 21:09, Chris Samuel wrote:

That said, there's been a lot of work done trying to catch issues which could leave the filesystem in an inconsistent state on loss of power (which was the most recent corruption issue I've noticed reported), which kernel version are you thinking of?

I take that back, the following 2 patches merged for 3.7 address potential filesystem corruption issues: commit 5af3e8cce8b7ba0a2819e18c9146c8c0b452d479 Author: Stefan Behrens <sbehrens@giantdisaster.de> Date: Wed Aug 1 18:56:49 2012 +0200 Btrfs: make filesystem read-only when submitting barrier fails So far the return code of barrier_all_devices() is ignored, which means that errors are ignored. The result can be a corrupt filesystem which is not consistent. This commit adds code to evaluate the return code of barrier_all_devices(). The normal btrfs_error() mechanism is used to switch the filesystem into read-only mode when errors are detected. In order to decide whether barrier_all_devices() should return error or success, the number of disks that are allowed to fail the barrier submission is calculated. This calculation accounts for the worst RAID level of metadata, system and data. If single, dup or RAID0 is in use, a single disk error is already considered to be fatal. Otherwise a single disk error is tolerated. The calculation of the number of disks that are tolerated to fail the barrier operation is performed when the filesystem gets mounted, when a balance operation is started and finished, and when devices are added or removed. Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de> commit 62856a9b73860cffe2a3d91b069393b88c219aa6 Author: Stefan Behrens <sbehrens@giantdisaster.de> Date: Tue Jul 31 11:09:44 2012 -0600 Btrfs: detect corrupted filesystem after write I/O errors In check-integrity, detect when a superblock is written that points to blocks that have not been written to disk due to I/O write errors. Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de> -- Chris Samuel : http://www.csamuel.org/ : Melbourne, VIC

On 04/11/12 14:33, Andrew McGlashan wrote:

Insofar as Oracle is concerned with their own version of RHEL.... BTRFS is production ready, but they haven't certified their other main products to use it!

I am very dubious of such claims, there should be no magic code in there that won't be in the mainline by now and Chris Mason left Oracle to join Fusion-IO back in June, and was joined by Josef Bacik from Red Hat. It's still marked experimental in the mainline kernel. # Btrfs is highly experimental, and THE DISK FORMAT IS NOT # YET FINALIZED. You should say N here unless you are interested # in testing Btrfs with non-critical data. cheers, Chris -- Chris Samuel : http://www.csamuel.org/ : Melbourne, VIC

Hi Avi, On 5/11/2012 8:04 AM, Avi Miller wrote:

On 04/11/2012, at 2:33 PM, Andrew McGlashan <andrew.mcglashan@affinityvision.com.au> wrote:

...

Insofar as Oracle is concerned with their own version of RHEL.... BTRFS is production ready, but they haven't certified their other main products to use it!

This is half-true: Oracle supports btrfs if used in production, but the Linux team is not responsible for product certification. So, you can happily store Oracle binaries on btrfs (and I know a few customers that are now exploring this), but you can't use btrfs to store Oracle Database data. And why would you anyway, given so many better options, not least of which is Oracle ASM?

I wish it was 0% true... ;)

And, Oracle Database is one of the few Oracle products that actually certifies filesystems for data storage. Where a product has no filesystem certifications at all, e.g. the middleware stack, you can also use btrfs.

I much prefer using standard files on a file system without autoextend enabled or any kind of automatic space management; it is much less prone to space issues when managed properly and it is much easier for me to clone databases for development and test requirements from a simple production backup. Cheers -- Kind Regards AndrewM

Hello Bruce, On Sat, 03 Nov 2012 13:38:28 +1100, Bruce Wang <bruce@brucewang.net> wrote:

Hi Terry,

Have been using SSD with Linux daily for over 3 months without any problems.

You might want to checkout this article http://wiki.debian.org/SSDoptimization

Thanks for that link, very helpful. Cheers, -- Regards, Terry Duell

Terry Duell <tduell@iinet.net.au> wrote:

I may have been a bit premature...the installed drive is an IDE.

I am very surprised that a new laptop would have an IDE drive these days. My laptop is over three years old and it has a SATA drive according to the hardware manual.

Terry Duell wrote:

On Sat, 03 Nov 2012 13:15:50 +1100, Terry Duell<tduell@iinet.net.au> wrote:

...

Hello All, To expand on the subject line, I have inherited an Acer laptop on which I intend to install Linux. I think the machine spec should be OK for most linux distros...Core 2 duo T5600, 4GB, 120GB, Nvidia graphics. I am considering replacing the 120GB drive with an SSD, as it should smarten up the performance a tad and probably improve battery life a bit as well.

Hi Terry, It is worth it from a performance point of view, although as you say, the IDE drives are more expensive. Depending on how much you plan to use the laptop CF may be an option, I have an old benq celeron mobile 1.6GHz/1GB ram running on a CF (8GB) and it speeds things up a lot in the boot department especially. And with CF at 32GB or greater at reasonable prices these days... Although if you plan to use it a bit I would still recommend SSD, as these are optimised for use as dynamic storage, I don't think the wear leveling on CF is as good.

I may have been a bit premature...the installed drive is an IDE. MSY only have SATA SSDs in their price list, and a quick snoop about suggests the IDE SSDs that are available are much more expensive. On price alone it may not be worth it, and I have yet to find info that will tell the story on the performance.

Cheers,

cheers Robert

I had a bit of a snoop about for info on CF, and it isn't obvious to me how I could use one on the Acer.

You used to be able to get IDE<->CF adapters that were the size of a 2.5" drive. I assume they are still around but maybe not so easy to get these days.

I'm now thinking I should just install Linux on the IDE drive and see how it performs.

That's probably a better starting point. What are the specs of your inherited laptop? I have an inherited HP IDE-era laptop that still performs pretty well. The Ethernet is fried on it, and the IDE disk sometimes fails to detect on boot. 1680x1050 resolution though - it would have been a pretty expensive unit when it was new. James

Terry Duell <tduell@iinet.net.au> wrote:

It's not a bad little unit, a bit slow starting Win XP, but will probably be better with a suitable linux. I'm still thinking about a distro, and have recently installed Mint 13 and Manjaro in Virtualbox on my main system, to get a feel.

I've read good reports about Arch Linux, but I haven't tried it. If you're concerned about performance, choose a distribution that lets you install a minimal operating system and then add packages to it. For example, Debian can be installed this way.

On Mon, Nov 5, 2012 at 8:44 AM, Terry Duell <tduell@iinet.net.au> wrote:

...

If you're concerned about performance, choose a distribution that lets you install a minimal operating system and then add packages to it. For example, Debian can be installed this way.

Yes, that would be a good approach. I have been looking at Mint and Mint-Debian, but really haven't spent enough time with any of them thus far.

I have a Panasonic toughbook I use for Ham radio in the field. It is a 1.4 GHZ processor and has 1.5GB of RAM and a 64GB IDE SSD, I am running Xubuntu on it and it runs quite fast, certainly much smoother then XP that came on it when I bought it. I cant say how much of a difference the SSD made as I swapped it in the same time I installed Linux. Xubuntu isnt as light as doing a minimal install of debian but it is much lighter then Mint and being Ubuntu based it all just worked for me including the touch screen, although I did have to make the touch screen calibration persistent across reboots. YMMV Mark -- Mark "Pockets" Clohesy Mob Phone: (+61) 406 417 877 Email: hiddensoul@twistedsouls.com G-Talk: mark.clohesy@gmail.com - GNU/Linux.. Linux Counter #457297 "I would love to change the world, but they won't give me the source code" "Linux is user friendly...its just selective about who its friends are" "Never underestimate the bandwidth of a V8 station wagon full of tapes hurtling down the highway" "The difference between e-mail and regular mail is that computers handle e-mail, and computers never decide to come to work one day and shoot all the other computers"

On 04/11/12 17:12, Jason White wrote:

I've read good reports about Arch Linux, but I haven't tried it.

I ran Arch on a 9 year old laptop with 512MB RAM when I found that none of the *buntu's install CD's would boot on it. It's firmly in the "configure it yourself" camp which I liked, and I ran it with LXDE for some time until the hardware started to be unstable. cheers, Chris -- Chris Samuel : http://www.csamuel.org/ : Melbourne, VIC

Terry Duell wrote:

[snip] Thanks for your help, but the SSD is now off the agenda, having found that the laptop has an IDE interface. Cheers, Pity it hasn't got a spare mini-PCIe slot; or you could try something like: http://ap.apacer.com/products/mPDM-M; which seems to be about 100GB; regards Rohan McLeod

On 05/11/12 10:07, Toby Corkindale wrote:

* set the 'discard' mount flag if you're using ext4, or 'ssd' on btrfs. (This tells the SSD when you've deleted stuff to free up space, and allows it to work a lot more efficiently)

That option is "discard" for btrfs too, the "ssd" option (which is automatically set if the kernel detects the disk as being non-rotational) determines allocation and commit log strategies. The "discard" option is not set by default as it can result in poor performance on some SSDs.

* set 'relatime' mount flag as well; it helps performance on everything, and in ssd cases, avoids a bunch of unnecessary writes.

That's the default on any vaguely modern kernel (set as default in 2.6.30 in 2009). :-) cheers, Chris -- Chris Samuel : http://www.csamuel.org/ : Melbourne, VIC

Arjen Lentz <arjen@lentz.com.au> wrote:

Unless you want to do auditing of read-access, atime is just not relevant. Incremental backups work on mtime not atime.

From memory, Mutt is one application that relies on atime (in that case, to determine which mail folders have been read since they were last updated). This still works if you use relatime though.

On 05/11/12 13:24, Arjen Lentz wrote:

Hi Chris, all

...

...

* set 'relatime' mount flag as well; it helps performance on everything, and in ssd cases, avoids a bunch of unnecessary writes.

That's the default on any vaguely modern kernel (set as default in 2.6.30 in 2009). :-)

"With this option enabled, atime data is written to the disk only if the file has been modified since the atime data was last updated (mtime), or if the file was last accessed more than a certain length of time ago (by default, one day)."

Unless you want to do auditing of read-access, atime is just not relevant. Incremental backups work on mtime not atime.

So I use noatime rather than relatime on my SSD laptops, just as we do on database partitions of servers since obviously there we fully expect there to be lots of reads and tracking that info is nonsense.

Agreed, I used noatime on my db volumes, and other systems where I know it'll be fine, but if I'm making blanket recommendations to strangers, relatime is safer.

Chris Samuel <chris@csamuel.org> writes:

On 05/11/12 10:07, Toby Corkindale wrote:

...

* set the 'discard' mount flag if you're using ext4, or 'ssd' on btrfs. (This tells the SSD when you've deleted stuff to free up space, and allows it to work a lot more efficiently)

That option is "discard" for btrfs too, the "ssd" option (which is automatically set if the kernel detects the disk as being non-rotational) determines allocation and commit log strategies.

Also ssd_sparse IIRC, for a different flavour of FTL. It's not very clear to me how to know which of the two to use.

Chris Samuel <chris@csamuel.org> writes:

On 09/11/12 11:07, Trent W. Buck wrote:

...

Also ssd_sparse IIRC, for a different flavour of FTL. It's not very clear to me how to know which of the two to use.

Aha, ssd_spread, looks like it's meant for lower grade SSD's.

What does that actually mean in practice, though? Once upon a time, it was intel, or <everything else>. Even if you can determine that your OCZ Vertex NN runs a Sandforce MM chipset, which do you use? Maybe the rule of thumb is to use ssd for SSDs, and ssd_spread for USB keys and SD/MMC cards?

samuel@eris:~/Downloads/linux$ git log -Sssd_spread fs/btrfs/ commit 451d7585a8bb1b9bec0d676ce3dece1923164e55 Author: Chris Mason <chris.mason@oracle.com> Date: Tue Jun 9 20:28:34 2009 -0400

Btrfs: add mount -o ssd_spread to spread allocations out

Some SSDs perform best when reusing block numbers often, while others perform much better when clustering strictly allocates big chunks of unused space.

The default mount -o ssd will find rough groupings of blocks where there are a bunch of free blocks that might have some allocated blocks mixed in.

mount -o ssd_spread will make sure there are no allocated blocks mixed in. It should perform better on lower end SSDs.

Signed-off-by: Chris Mason <chris.mason@oracle.com>