ZFS in OS X 10.5?
toby
- http://www.osnews.com/story.php/16739/Screenshot-ZFS-in-Leopard
- http://mac4ever.com/news/27485/zettabyte_sur_leopard/
(older rumour http://www.osnews.com/story.php?news_id=14473)
For OS X people wondering why the fuss about ZFS - summaries include:
- http://www.sun.com/2004-0914/feature/
- http://www.sun.com/bigadmin/features/articles/zfs_part1.scalable.html
"Why ZFS for home":
- http://uadmin.blogspot.com/2006/05/why-zfs-for-home.html
"Here are ten reasons why you'll want to reformat all of your systems
and use ZFS.":
http://www.tech-recipes.com/rx/1446/zfs_ten_reasons_to_reformat_your_hard_drives
And some more technical explanations from Chief Engineer:
- http://blogs.sun.com/bonwick/entry/zfs_end_to_end_data
- http://blogs.sun.com/bonwick/entry/smokin_mirrors
Jon
> Well now, isn't this interesting? (Not to mention thrilling):
Is it possible that this is (part of) the mechanism behind "Time
Machine", the mechanism for painless rollback to a previous file system
state that Leopard boasts, according to released information?
Quote from "Why ZFS for Home" (see previous post for URL):
"...another problem [than disk failure] is accidental deletion,
accidentally installing a broken application or any change you would
like to undo. Linux's answer to this is backups, either on optical
media, tape or perhaps another harddrive. This is expensive or time
consuming, choose one. So the typical home user will most likely put
this off to till another day and won't have a backup for their data.
ZFS has snapshots, they are easy and painless and have a very low cost
in resources to create. Snapshots are basically a picture of your data;
these are taken in real time and are nearly instant in ZFS, to get these
in any other OS you need to buy expensive raid hardware or an expensive
software package something that no home user will want to buy."
Ve-e-ry interesting.
--
/Jon
For contact info, run the following in Terminal:
Mail: echo 36199371860304980107073482417748002696458P|dc
Skype: echo 139576319600233690471689738P|dc
Chris Ridd
> Well now, isn't this interesting? (Not to mention thrilling):
> - http://www.osnews.com/story.php/16739/Screenshot-ZFS-in-Leopard
> - http://mac4ever.com/news/27485/zettabyte_sur_leopard/
That might just be someone tooling around with Interface Builder. I
mean - a 160MB disk??
I *hope* the rumours are correct, but I don't think this reinforces
them or not.
Cheers,
Chris
Chris Ridd
> toby <to...@telegraphics.com.au> wrote:
>
>> Well now, isn't this interesting? (Not to mention thrilling):
>
> Is it possible that this is (part of) the mechanism behind "Time
> Machine", the mechanism for painless rollback to a previous file system
> state that Leopard boasts, according to released information?
Possibly, and possibly not. Read
<http://arstechnica.com/staff/fatbits.ars/2006/8/15/4995>
Cheers,
Chris
toby
> On 2006-12-17 06:49:05 +0000, "toby" <to...@telegraphics.com.au> said:
>
> > Well now, isn't this interesting? (Not to mention thrilling):
> > - http://www.osnews.com/story.php/16739/Screenshot-ZFS-in-Leopard
> > - http://mac4ever.com/news/27485/zettabyte_sur_leopard/
>
> That might just be someone tooling around with Interface Builder. I
> mean - a 160MB disk??
Good point. Funny if that turned out to be a UI bug, though :)
>
> I *hope* the rumours are correct, but I don't think this reinforces
> them or not.
I tend to agree - but hope too. Steve has really kept the lid on this
one, but he's savvy enough to have green lighted this :)
>
> Cheers,
>
> Chris
Kevin McMurtrie
"toby" <to...@telegraphics.com.au> wrote:
OS 10.4.8 still has bugs that corrupt HFS+ filesystems, especially on
the quad core Intels. It wasn't until late in the 10.3.x releases that
HFS+ was reliable on even single PowerPC Macs.
I wonder how buggy Apple's ZFS filesystem will be?
toby
That's rubbish, at least in respect to PowerPC. (Can't speak for Intel
Macs, I don't use them.)
>
> I wonder how buggy Apple's ZFS filesystem will be?
You mean Sun's? Obviously a filesystem demands extraordinary attention
to QA. You have seen from Sun's articles on the subject that it has
already received a remarkable amount of testing within Sun. Assuming
the rumours are correct - Apple has clearly had some time to work on
integration with OS X. As of today it probably has plenty of bugs,
that's why software doesn't arrive on the day it's announced...
But this is all beside the point. If the rumours turn out to be true,
we have reason to celebrate - it will instantly put Apple a generation
ahead in storage management. From a quality standpoint I'd much prefer
they integrate ZFS than try to duplicate its functionality themselves.
Ian Collins
> Well now, isn't this interesting? (Not to mention thrilling):
> - http://www.osnews.com/story.php/16739/Screenshot-ZFS-in-Leopard
> - http://mac4ever.com/news/27485/zettabyte_sur_leopard/
> (older rumour http://www.osnews.com/story.php?news_id=14473)
>
--
Ian Collins.
Neill Massello
> Is it possible that this is (part of) the mechanism behind "Time
> Machine", the mechanism for painless rollback to a previous file system
> state that Leopard boasts, according to released information?
No. See <http://arstechnica.com/staff/fatbits.ars/2006/8/15/4995/p2>:
"Time Machine is not an interface to file system snapshots built on any
sort of new, modern file system. Instead, it's an automated backup
system that works with plain old HFS+. The point-in-time views in Time
Machine are actually sparsely populated directory trees on an external
disk or server containing mostly hard links to unchanged directories,
plus full copies of the few files that have been created or modified
since the last backup."
Jon
> No. See <http://arstechnica.com/staff/fatbits.ars/2006/8/15/4995/p2>:
Yes. I read that after posting.
Sad. because it would have been great, IMHO.
Jon
> Kevin McMurtrie wrote:
> > OS 10.4.8 still has bugs that corrupt HFS+ filesystems, especially on
> > the quad core Intels. It wasn't until late in the 10.3.x releases that
> > HFS+ was reliable on even single PowerPC Macs.
>
> That's rubbish, at least in respect to PowerPC. (Can't speak for Intel
> Macs, I don't use them.)
At least you (Kevin) need to specify. What bugs?
Neill Massello
> Sad. because it would have been great, IMHO.
After all the changes that Mac users and developers have been through in
the past few years, a new default file system right now would probably
be too much. Time Machine is being pitched as an easy backup method for
ordinary users, so it needs to work with the current default file system
and not require any reformatting or reinstalling. Before Apple makes ZFS
the default, or even a requirement for some optional feature, they'll
want to provide for a generous period of preparation and testing by
outside developers and adventurous users. Support for it in 10.5 might
be a step toward its adoption as the new Mac OS file system.
Jolly Roger
> Neill Massello <mass...@newsguy.com> wrote:
>
>> No. See <http://arstechnica.com/staff/fatbits.ars/2006/8/15/4995/p2>:
>
> Yes. I read that after posting. Sad. because it would have been great, IMHO.
Speak for yourself. I, for one, am not interested in having to
reformat and reinstall on all 10 of my Macs just to be able to use Time
Machine!
--
JR
toby
Snapshots are only one feature. You might find the Sun material above
interesting for details of the rest of the story. Think of this as OS X
coming up to speed with the state of the art - Isn't that a good thing,
even if you're not ready to "reformat" yourself?
>
> --
> JR
Paul Sture
"toby" <to...@telegraphics.com.au> wrote:
> Kevin McMurtrie wrote:
> >
> > I wonder how buggy Apple's ZFS filesystem will be?
>
> You mean Sun's? Obviously a filesystem demands extraordinary attention
> to QA. You have seen from Sun's articles on the subject that it has
> already received a remarkable amount of testing within Sun. Assuming
> the rumours are correct - Apple has clearly had some time to work on
> integration with OS X. As of today it probably has plenty of bugs,
> that's why software doesn't arrive on the day it's announced...
You are dead right about QA there. Were I a betting man I'd put good
money on the fact that some customer somewhere will have just the right
mix of hardware, software and workload to discover problems with any new
file system.
> But this is all beside the point. If the rumours turn out to be true,
> we have reason to celebrate - it will instantly put Apple a generation
> ahead in storage management. From a quality standpoint I'd much prefer
> they integrate ZFS than try to duplicate its functionality themselves.
I honestly doubt that Apple has anywhere near Sun's expertise in this
area.
--
Paul Sture
toby
Which can only be a good argument for using Sun's technology, eh.
>
> --
> Paul Sture
Paul Sture
"toby" <to...@telegraphics.com.au> wrote:
Subject to the right agreements between Apple and Sun being reached,
yes. Of course, putting that technology onto a different flavour of Unix
introduces another phase of QA.
--
Paul Sture
Gregory Weston
<paul.sture.nospam-4...@mac.sture.homeip.net>,
Paul Sture <paul.stu...@hispeed.ch> wrote:
> > But this is all beside the point. If the rumours turn out to be true,
> > we have reason to celebrate - it will instantly put Apple a generation
> > ahead in storage management. From a quality standpoint I'd much prefer
> > they integrate ZFS than try to duplicate its functionality themselves.
>
> I honestly doubt that Apple has anywhere near Sun's expertise in this
> area.
Based on what? Apple's got some fairly formidable people in their file
system group.
--
The best intentions in the world don't make a flawed argument magically valid.
Steinar Moum
sn...@spambin.fsnet.co.uk (Sn!pe) wrote:
> Gregory Weston <u...@splook.com> wrote:
> Could someone please explain to this know-nothing what
> the advantages of ZFS are?
You will find informative information if you do a google search for ZFS.
sm
Gregory Weston
sn...@spambin.fsnet.co.uk (Sn!pe) wrote:
> Gregory Weston <u...@splook.com> wrote:
>
> > > > But this is all beside the point. If the rumours turn out to be true,
> > > > we have reason to celebrate - it will instantly put Apple a generation
> > > > ahead in storage management. From a quality standpoint I'd much prefer
> > > > they integrate ZFS than try to duplicate its functionality themselves.
> > >
> > > I honestly doubt that Apple has anywhere near Sun's expertise in this
> > > area.
> >
> > Based on what? Apple's got some fairly formidable people in their file
> > system group.
>
> Could someone please explain to this know-nothing what
> the advantages of ZFS are?
Some of the links from the original post are useful:
For OS X people wondering why the fuss about ZFS - summaries include:
- http://www.sun.com/2004-0914/feature/
- http://www.sun.com/bigadmin/features/articles/zfs_part1.scalable.html
"Why ZFS for home":
- http://uadmin.blogspot.com/2006/05/why-zfs-for-home.html
"Here are ten reasons why you'll want to reformat all of your systems
and use ZFS.":
http://www.tech-recipes.com/rx/1446/zfs_ten_reasons_to_reformat_your_hard
_drives
And some more technical explanations from Chief Engineer:
- http://blogs.sun.com/bonwick/entry/zfs_end_to_end_data
- http://blogs.sun.com/bonwick/entry/smokin_mirrors
--
Paul Sture
Gregory Weston <u...@splook.com> wrote:
> In article
> <paul.sture.nospam-4...@mac.sture.homeip.net>,
> Paul Sture <paul.stu...@hispeed.ch> wrote:
>
> > > But this is all beside the point. If the rumours turn out to be true,
> > > we have reason to celebrate - it will instantly put Apple a generation
> > > ahead in storage management. From a quality standpoint I'd much prefer
> > > they integrate ZFS than try to duplicate its functionality themselves.
> >
> > I honestly doubt that Apple has anywhere near Sun's expertise in this
> > area.
>
> Based on what? Apple's got some fairly formidable people in their file
> system group.
I'm not saying Apple haven't got excellent people; rather that Sun has a
lot of experience with customers who process extremely high volumes of
data (and spend megabucks doing so).
--
Paul Sture
Rich Teer
> Subject to the right agreements between Apple and Sun being reached,
That won't be necessary: ZFS is open source, so Apple are free to
use it if they wish. Which would make more sense than trying to
reimplement it from scratch.
> yes. Of course, putting that technology onto a different flavour of Unix
> introduces another phase of QA.
Agreed.
--
Rich Teer, SCNA, SCSA, SCSECA, OpenSolaris CAB member
. * * . * .* .
. * . .*
President, * . . /\ ( . . *
Rite Online Inc. . . / .\ . * .
.*. / * \ . .
. /* o \ .
Voice: +1 (250) 979-1638 * '''||''' .
URL: http://www.rite-group.com/rich ******************
Jolly Roger
> Jolly Roger wrote:
>>
>> Speak for yourself. I, for one, am not interested in having to
>> reformat and reinstall on all 10 of my Macs just to be able to use Time
>> Machine!
>
> Snapshots are only one feature. You might find the Sun material above
> interesting for details of the rest of the story. Think of this as OS X
> coming up to speed with the state of the art - Isn't that a good thing,
> even if you're not ready to "reformat" yourself?
Sure, ZFS is good and all, but I'd rather Time Machine not be dependent on it.
--
JR
Jolly Roger
>> OS 10.4.8 still has bugs that corrupt HFS+ filesystems, especially on
>> the quad core Intels. It wasn't until late in the 10.3.x releases that
>> HFS+ was reliable on even single PowerPC Macs.
>
> That's rubbish, at least in respect to PowerPC. (Can't speak for Intel
> Macs, I don't use them.)
I think I know what Kevin is talking about (correct me if I'm wrong, Kevin).
There's reportedly a long-standing problem where if you fill a HFS
drive to more than 85% capacity with many files, the file system may
overwrite extents causing directory corruption. More info here:
This
problem reportedly existed at least up to 1999, and may still exist today.
--
JR
Jon
> You will find informative information if you do a google search for ZFS.
Or go back to the beginning of the thread and look at the many links
proided in the OP.
toby
Gregory Weston wrote:
> In article
> <paul.sture.nospam-4...@mac.sture.homeip.net>,
> Paul Sture <paul.stu...@hispeed.ch> wrote:
>
> > > But this is all beside the point. If the rumours turn out to be true,
> > > we have reason to celebrate - it will instantly put Apple a generation
> > > ahead in storage management. From a quality standpoint I'd much prefer
> > > they integrate ZFS than try to duplicate its functionality themselves.
> >
> > I honestly doubt that Apple has anywhere near Sun's expertise in this
> > area.
>
> Based on what? Apple's got some fairly formidable people in their file
> system group.
True. But Sun beat them to it. People should really look at what ZFS is
before assuming it's "just another filesystem".
ZnU
Jolly Roger <jolly...@null.org> wrote:
According to various reports, the version presently in Leopard isn't.
The early use case for ZFS is going to be big storage arrays, not your
internal boot disk. Supposedly it isn't even bootable yet. Maybe down
the road Apple will switch to it as the default. (Assuming any of this
is accurate, of course.)
It does certainly have wide potential applications on the desktop, even
in single-drive systems. Just as one example... sometimes there's a
concern that installing something or running some sort of utility will
screw up the system (it doesn't happen much in OS X, but it's not
entirely unheard of). Snapshots would make this kind of thing safe; if
things go wrong, you can revert the entire system to a previous state.
(Time Machine can't, as far as we know, operate system-wide like that.)
--
"Those who enter the country illegally violate the law."
-- George W. Bush in Tucson, Ariz., Nov. 28, 2005
Steve Hix
sn...@spambin.fsnet.co.uk (Sn!pe) wrote:
> Gregory Weston <u...@splook.com> wrote:
>
> > > > But this is all beside the point. If the rumours turn out to be true,
> > > > we have reason to celebrate - it will instantly put Apple a generation
> > > > ahead in storage management. From a quality standpoint I'd much prefer
> > > > they integrate ZFS than try to duplicate its functionality themselves.
> > >
> > > I honestly doubt that Apple has anywhere near Sun's expertise in this
> > > area.
> >
> > Based on what? Apple's got some fairly formidable people in their file
> > system group.
>
> Could someone please explain to this know-nothing what
> the advantages of ZFS are?
This is a bit dated, but it's a start:
Steve Hix
Gregory Weston <u...@splook.com> wrote:
[snip]
>
> "Here are ten reasons why you'll want to reformat all of your systems
> and use ZFS.":
> http://www.tech-recipes.com/rx/1446/zfs_ten_reasons_to_reformat_your_hard
> _drives
Except for the minor problem of actually booting your system. (Even
under Solaris, so far.)
But for attached non-boot volumes, it's pretty slick.
Steve Hix
It isn't. It depends on yet more extensions tacked on to HFS+.
But in the future...
Gregory Weston
"toby" <to...@telegraphics.com.au> wrote:
> Gregory Weston wrote:
> > In article
> > <paul.sture.nospam-4...@mac.sture.homeip.net>,
> > Paul Sture <paul.stu...@hispeed.ch> wrote:
> >
> > > > But this is all beside the point. If the rumours turn out to be true,
> > > > we have reason to celebrate - it will instantly put Apple a generation
> > > > ahead in storage management. From a quality standpoint I'd much prefer
> > > > they integrate ZFS than try to duplicate its functionality themselves.
> > >
> > > I honestly doubt that Apple has anywhere near Sun's expertise in this
> > > area.
> >
> > Based on what? Apple's got some fairly formidable people in their file
> > system group.
>
> True. But Sun beat them to it. People should really look at what ZFS is
> before assuming it's "just another filesystem".
I'm not aware that anyone in this thread has made that assumption.
Especially since the OP kindly pointed to some documentation and
commentary on it.
If there's meaning to the fact that you chose to use my post as the
basis for your own comment, you should realize that I meant nothing more
than what I said. (And Paul later clarified what he meant by "this
area.")
ZnU
Steve Hix <se...@NOSPAMspeakeasy.netINVALID> wrote:
Fairly bizarre ones, no less. (One thing Apple added to HFS+ to support
Time Machine is hard links to directories, which don't exist in any
other system, as far as I'm aware.)
It almost looks like they were planning Time Machine with ZFS in mind,
and then decided it needed to work on HFS+ as well, so they hacked
something up to make that work. (Maybe just as a convenience for people
who don't want to switch file systems, maybe because ZFS won't be ready
for use as the default file system in Leopard yet.)
> But in the future...
Neill Massello
> It almost looks like they were planning Time Machine with ZFS in mind,
> and then decided it needed to work on HFS+ as well, so they hacked
> something up to make that work. (Maybe just as a convenience for people
> who don't want to switch file systems, maybe because ZFS won't be ready
> for use as the default file system in Leopard yet.)
Even if it is ready, making Time Machine dependent on it would be a
marketing mistake, as TM is being touted as an easy backup solution for
precisely the kind of users who are least likely to want to
re-initialize their drives. I don't think Apple will switch to ZFS as
the default until it's been publicly available for a while. (I assume
that all of Adobe's apps now work on case-sensitive volumes?)
Kevin McMurtrie
> Kevin McMurtrie wrote:
> > In article <1166338145.7...@n67g2000cwd.googlegroups.com>,
> > "toby" <to...@telegraphics.com.au> wrote:
> >
> > > Well now, isn't this interesting? (Not to mention thrilling):
> > > - http://www.osnews.com/story.php/16739/Screenshot-ZFS-in-Leopard
> > > - http://mac4ever.com/news/27485/zettabyte_sur_leopard/
> > > (older rumour http://www.osnews.com/story.php?news_id=14473)
> > >
> > > For OS X people wondering why the fuss about ZFS - summaries include:
> > > - http://www.sun.com/2004-0914/feature/
> > > - http://www.sun.com/bigadmin/features/articles/zfs_part1.scalable.html
> > >
> > > "Why ZFS for home":
> > > - http://uadmin.blogspot.com/2006/05/why-zfs-for-home.html
> > >
> > > "Here are ten reasons why you'll want to reformat all of your systems
> > > and use ZFS.":
> > > http://www.tech-recipes.com/rx/1446/zfs_ten_reasons_to_reformat_your_hard_
> > > driv
> > > es
> > >
> > > And some more technical explanations from Chief Engineer:
> > > - http://blogs.sun.com/bonwick/entry/zfs_end_to_end_data
> > > - http://blogs.sun.com/bonwick/entry/smokin_mirrors
> >
> > OS 10.4.8 still has bugs that corrupt HFS+ filesystems, especially on
> > the quad core Intels. It wasn't until late in the 10.3.x releases that
> > HFS+ was reliable on even single PowerPC Macs.
>
> That's rubbish, at least in respect to PowerPC. (Can't speak for Intel
> Macs, I don't use them.)
Maybe you're not driving your system very hard. As a developer, I have
over 600000 files on my hard drive and software may be using tens of
thousands of files actively. 10.4 was the first OS where filesystem
corruption stopped being a regular problem. With 10.3, booting into
single user mode to run fsck was part of the Monday morning routine at
home and work.
My new quad core Intel at work has corrupted its HFS+ filesystem twice.
The /var/vm/app_profile database gets trashed quite a bit too and the
results are ugly. I guess not all of the multi-threading/multi-CPU bugs
have been fixed.
> >
> > I wonder how buggy Apple's ZFS filesystem will be?
>
> You mean Sun's? Obviously a filesystem demands extraordinary attention
> to QA. You have seen from Sun's articles on the subject that it has
> already received a remarkable amount of testing within Sun. Assuming
> the rumours are correct - Apple has clearly had some time to work on
> integration with OS X. As of today it probably has plenty of bugs,
> that's why software doesn't arrive on the day it's announced...
>
> But this is all beside the point. If the rumours turn out to be true,
> we have reason to celebrate - it will instantly put Apple a generation
> ahead in storage management. From a quality standpoint I'd much prefer
> they integrate ZFS than try to duplicate its functionality themselves.
I'd jump for joy if Apple debugged CIFS compatibility or supported
WebDav for real. Is there a network protocol for ZFS yet?
Kevin McMurtrie
see_si...@mac.com.invalid (Jon) wrote:
> toby <to...@telegraphics.com.au> wrote:
> > Kevin McMurtrie wrote:
> > > OS 10.4.8 still has bugs that corrupt HFS+ filesystems, especially on
> > > the quad core Intels. It wasn't until late in the 10.3.x releases that
> > > HFS+ was reliable on even single PowerPC Macs.
> >
> > That's rubbish, at least in respect to PowerPC. (Can't speak for Intel
> > Macs, I don't use them.)
>
> At least you (Kevin) need to specify. What bugs?
My new quad core Intel got an I/O error when I deleted a file. The
system log said something like, "Block not allocated." I ran fsck in
single user mode and there were several fresh problems with the volume.
The VM profiling database in /var/vm/app_profile has become corrupted
three times so far, which causes apps to consistently crash.
My PowerPC systems usually had problems with preference files vanishing
or becoming unwritable even through nothing produced an error. Letting
fsck fix the volume took care of it. They've been pretty much OK under
10.4.
Gregory Weston
<mcmurtri-444ABD...@sn-radius.vsrv-sjc.supernews.net>,
Kevin McMurtrie <mcmu...@dslextreme.com> wrote:
> > That's rubbish, at least in respect to PowerPC. (Can't speak for Intel
> > Macs, I don't use them.)
>
> Maybe you're not driving your system very hard. As a developer, I have
> over 600000 files on my hard drive and software may be using tens of
> thousands of files actively. 10.4 was the first OS where filesystem
> corruption stopped being a regular problem. With 10.3, booting into
> single user mode to run fsck was part of the Monday morning routine at
> home and work.
If corruption was a "regular problem" for you such that you were
repairing your disks weekly, there was something very atypical about
your installation.
G
Hugh Watkins
which sounds like a job for the parallels transporter Beta
making a complete vm bootable disk image
Hugh W
--
a wonderful artist in Denmark
http://www.ingerlisekristoffersen.dk/
Beta blogger
http://snaps4.blogspot.com/ photographs and walks
old blogger GENEALOGE
http://hughw36.blogspot.com/ MAIN BLOG
toby
I don't know what you're doing differently, but I've *never* seen this
problem on any version of MacOS. I have been running OS X alone for at
least 50-100 machine-years. And I'm a developer.
It's not HFS+, that's certain.
>
> My new quad core Intel at work has corrupted its HFS+ filesystem twice.
> The /var/vm/app_profile database gets trashed quite a bit too and the
> results are ugly. I guess not all of the multi-threading/multi-CPU bugs
> have been fixed.
As I said above, I can't speak for Intel, but on PowerPC it's quite
stable.
Rich Teer
> I'd jump for joy if Apple debugged CIFS compatibility or supported
> WebDav for real. Is there a network protocol for ZFS yet?
Yeah, it's called NFS...
Chris Ridd
> On Mon, 18 Dec 2006, Kevin McMurtrie wrote:
>
>> I'd jump for joy if Apple debugged CIFS compatibility or supported
>> WebDav for real. Is there a network protocol for ZFS yet?
>
> Yeah, it's called NFS...
There's also iSCSI target support for ZFS in OpenSolaris build 53, if
that's your thing.
<http://www.cuddletech.com/blog/pivot/entry.php?id=775>
Cheers,
Chris
ZnU
"toby" <to...@telegraphics.com.au> wrote:
I've also never had file system corruption on OS X, and I've been
running it since the beginning. I've presently got 570K files totaling
635 GB on an internal RAID (which I also boot from). HFS+ seems to hold
up just fine.
[snip]
toby
I'd go further. By far the most likely cause of corruption in highly
mature filesystems (ext2fs, HFS, HFS+, UFS, etc) is a hardware problem.
And ZFS is in a better position to detect and recover from such
problems (cable, controller, drive, memory, cache) than any other
system I'm aware of. So whether you believe HFS+ to be buggy, or
whether there's some other factor, you're going to be better off with
ZFS.
Also, with regression tests and data from Apple's BTS, the reliability
of HFS+ can be fairly accurately estimated. You'd better believe Apple
does a LOT of such tests.
Daniel Packman
Kevin McMurtrie <mcmu...@dslextreme.com> wrote:
....
>My new quad core Intel got an I/O error when I deleted a file. The
>system log said something like, "Block not allocated."
This seems most likely to be a hardware problem, specifically and media
failure on a given block in the disk.
> I ran fsck in
>single user mode and there were several fresh problems with the volume.
>The VM profiling database in /var/vm/app_profile has become corrupted
>three times so far, which causes apps to consistently crash.
Again, this seems most likely to be a hardware problem.
>My PowerPC systems usually had problems with preference files vanishing
>or becoming unwritable even through nothing produced an error. Letting
>fsck fix the volume took care of it. They've been pretty much OK under
>10.4.
Now *this* might be related to bugs and/or software design issues.
A potential problem with all but the newest Macintosh machines is bit
errors from non-ecc memory.
toby
Kevin McMurtrie wrote:
> In article <1hqigmo.hpeb371yv6nsrN%see_si...@mac.com.invalid>,
> see_si...@mac.com.invalid (Jon) wrote:
>
> > toby <to...@telegraphics.com.au> wrote:
> > > Kevin McMurtrie wrote:
> > > > OS 10.4.8 still has bugs that corrupt HFS+ filesystems, especially on
> > > > the quad core Intels. It wasn't until late in the 10.3.x releases that
> > > > HFS+ was reliable on even single PowerPC Macs.
> > >
> > > That's rubbish, at least in respect to PowerPC. (Can't speak for Intel
> > > Macs, I don't use them.)
> >
> > At least you (Kevin) need to specify. What bugs?
>
> My new quad core Intel got an I/O error when I deleted a file. The
> system log said something like, "Block not allocated." I ran fsck in
> single user mode and there were several fresh problems with the volume.
> The VM profiling database in /var/vm/app_profile has become corrupted
> three times so far, which causes apps to consistently crash.
Thanks for the detail. Have you tested your RAM?
Paul Sture
"toby" <to...@telegraphics.com.au> wrote:
> And ZFS is in a better position to detect and recover from such
> problems (cable, controller, drive, memory, cache) than any other
> system I'm aware of.
That's a very valid point, and one I have come across in tape backup
discussions. Yes, you can save CPU cycles by relying on CRC provided by
tape drive hardware, but a CRC in the backup software itself covers the
full data path to the tape drive.
--
Paul Sture
toby
A lot of people don't get this... yet. Which is symptomatic of the
generational shift it represents. Sun (or more likely Apple) would do
well to create a "ZFS for dummies" presentation.
The OpenSolaris site does have some basic "screencasts", though:
http://www.opensolaris.org/os/community/zfs/demos/
>
> --
> Paul Sture
David C.
>
> A potential problem with all but the newest Macintosh machines is bit
> errors from non-ecc memory.
Correction. With almost all computers.
Consumer PC's haven't used parity memory for a very long time. And it
is rare to find a non-server PC that can support ECC memory.
-- David
Paul Sture
"toby" <to...@telegraphics.com.au> wrote:
Thanks for the links.
--
Paul Sture
Jon
> Well now, isn't this interesting? (Not to mention thrilling):
Yes, it is! :-)
And ThinkSecret reports today:
<http://www.thinksecret.com/news/0612leopard9a321gallery.html>
"New to build 9A321 is support for Sun's ZFS file system, a 128-bit open
source file system introduced with Solaris 10 that offers support for
vastly larger drives and arrays than 64-bit file systems. ZFS also
delivers additional options for administrators." (Now there is an
understatement...)
However, it also states that Leopard still has a long way to go before
release, effectively killing hope of an imminent release around MWSF.
toby
Jon wrote:
> toby <to...@telegraphics.com.au> wrote:
>
> > Well now, isn't this interesting? (Not to mention thrilling):
>
> Yes, it is! :-)
>
> And ThinkSecret reports today:
> <http://www.thinksecret.com/news/0612leopard9a321gallery.html>
>
> "New to build 9A321 is support for Sun's ZFS file system, a 128-bit open
> source file system introduced with Solaris 10 that offers support for
> vastly larger drives and arrays than 64-bit file systems.
Which is to most people, the least important thing... Feh. As I said in
another post, let's hope, after an official announcement, Apple does a
great job of communicating ZFS' real significance.
> ZFS also
> delivers additional options for administrators." (Now there is an
> understatement...)
>
> However, it also states that Leopard still has a long way to go before
> release,
More QA is always good :)
> effectively killing hope of an imminent release around MWSF.
> --
> /Jon
> For contact info, run the following in Terminal:
> Mail: echo 36199371860304980107073482417748002696458P|dc
> Skype: echo 139576319600233690471689738P|dc
This is the cleverest sig block I've seen in a while! (Assuming you
don't want to hear from Windows users, haha.)
Marc
>> For contact info, run the following in Terminal:
>> Mail: echo 36199371860304980107073482417748002696458P|dc
>> Skype: echo 139576319600233690471689738P|dc
>
> This is the cleverest sig block I've seen in a while! (Assuming you
> don't want to hear from Windows users, haha.)
Or solaris users. On solaris (both sparc and x86) it outputs garbage.
toby
Yep. Assumes GNU dc. Even more selective :)
Eric Lindsay
"toby" <to...@telegraphics.com.au> wrote:
> Jon wrote:
> > /Jon
> > For contact info, run the following in Terminal:
> > Mail: echo 36199371860304980107073482417748002696458P|dc
> > Skype: echo 139576319600233690471689738P|dc
>
> This is the cleverest sig block I've seen in a while! (Assuming you
> don't want to hear from Windows users, haha.)
First time I saw that I wondered wtf Jon had done, and how he did it.
Then man dc brought even more confusion (read and see). Maybe Jon will
explain how he generates it. I got help with:
perl -e 'print"ibase=16\n".(uc
unpack"H*","eric.me\@wherever.net.au\n")."\n";' | bc
but I don't know if that is the usual approach.
Oh yes, and ZFS does look an interesting development for OSX, if it
appears. Might be a longer wait than expected for Leopard. Maybe I won't
want for Leopard before buying an Intel Mac.
Jon
> Maybe Jon will
> explain how he generates it.
I think that this URL combined with a little head-scratching should do
the trick:
<http://synflood.at/blog/index.php?/archives/483-dc1-magic-auto-generati
on.html>
Happy holidays!
--
Jon
> In article <1hqoxgz.5ls9us1sxhvwxN%see_si...@mac.com.invalid>,
> see_si...@mac.com.invalid (Jon) wrote:
>
> > However, it also states that Leopard still has a long way to go
> > before release, effectively killing hope of an imminent release
> > around MWSF.
>
> I don't think that anyone had any realistic expectation of that
> happening.
Neither do I, but there are always some...
I would much, much rather have a reasonably well tested 10.5.0 release
than a hurry-to-market "beta" release like 10.4.
Which is also why I think it is nice that ZFS is now (presumably) being
offered as an option, not as default, at least not for another release
or two.
Rich Teer
> /Jon
> For contact info, run the following in Terminal:
> Mail: echo 36199371860304980107073482417748002696458P|dc
Hmm:
rich@marrakesh4103# echo 36199371860304980107073482417748002696458P|dc
:@EPJ0IF
P1<%c=rich@marrakesh4104#
I take it you don't want people to contact you... ;-)
Jon
> I take it you don't want people to contact you... ;-)
Well, not from usenet unless they take the trouble to do this, no. ;-)
And someone told me this doesn't work in Solaris. I don't know why,
sorry...
--
/Jon
For contact info, run the following in Terminal:
Mail: echo 36199371860304980107073482417748002696458P|dc
Skype: echo 139576319600233690471689738P|dc
Chris Ridd
> Rich Teer <rich...@rite-group.com> wrote:
>
>> I take it you don't want people to contact you... ;-)
>
> Well, not from usenet unless they take the trouble to do this, no. ;-)
> And someone told me this doesn't work in Solaris. I don't know why,
> sorry...
It only works with GNU versions of dc. If you install that on Solaris,
it works fine.
Cheers,
Chris
Jon
> It only works with GNU versions of dc. If you install that on Solaris,
> it works fine.
Ah. There it was. I knew someone told me once. Thanks!
And happy holidays, everyone!
Eric Lindsay
see_si...@mac.com.invalid (Jon) wrote:
> Rich Teer <rich...@rite-group.com> wrote:
>
> > I take it you don't want people to contact you... ;-)
>
> Well, not from usenet unless they take the trouble to do this, no. ;-)
> And someone told me this doesn't work in Solaris. I don't know why,
> sorry...
Probably the Solaris version lacks something assumed by this peculiar
part of the dc port.
P Pops off the value on top of the stack. If it it a string, it is
simply printed without a trailing newline. Otherwise it is a number,
and the integer portion of its absolute value is printed out as a
"base (UCHAR_MAX+1)" byte stream. Assuming that (UCHAR_MAX+1) is
256 (as it is on most machines with 8-bit bytes), the sequence
KSK 0k1/ [_1*]sx d0>x [256~aPd0<x]dsxxsxLKk could also accomplish this
function, except for the side-effect of clobbering the x register.
Jon
> Probably the Solaris version lacks something assumed by this peculiar
> part of the dc port.
See Chris' posting, where he says:
> It only works with GNU versions of dc. If you install that on
> Solaris, it works fine.