virtual disk=>full FILE-based restore,problem

I'm a sysadmin & developer with a lot of experience in
performance tuning. My philosophy for defragging: make a
backup of a *single* partition, reformat the partition, then do
folder-by-folder restores to control exactly where I want each
file to be on the disk.

Since TI is SECTOR-based, the only way to do this with a TI
backup is to mount the image as a virtual disk, and then use
non-Acronis utilities (e.g., Windows' native Explorer) to
accomplish the copying of individual folders to the
re-formatted partition. Knowing what I do about low-level disk
architecture, I can see no reason why this shouldn't work.

But when I do this I have two problems:
(1) the restored partition will boot all the way through login,
but it's impossible get explorer to run -- no taskbar or desktop
icons, etc. I've tried tricks using ctrl-alt-del=>Taskman,
but explorer quits instantly after starting.
(2) BSOD (kmode-exception-not-handled, mup.sys) during shutdown.

Why shouldn't this work?

 

I'm not sure, but it could be, if you restore the drive in full (so sector based restore) that Acronis rearranges the used sectors? It has to anyway if the drive geometry is different on the destination.

I used to "defrag" like that using Ghost 2003, and the result was always a faster running windows. I've never really actually checked with defrag...

 

Hi A-C,

How many partitions do you have?
Which partition is your boot partition?
Where is Windows installed?
Which partition are you "defragmenting" in this way?
Are you defragmentation the same partition that is your active boot/Windows partition?

 

Hello A-C,

Thank you for choosing Acronis Disk Backup Software.

Please note that Acronis True Image can change the destination of some information that is not sector-bound. It is done to improve the performance and the imaging/restoring speed. In fact, you get the fragmented disk where only some data (sector-bound) is on the exactly same sectors as it was before. This seems to be the reason for the problem you described.

Thank you.
--
Ilya Toytman

 

Combined reply to MiniMax & Acronis Support...

MiniMax: I have 9 partitions, each is barely less than 8GB.
I boot w2kPro, mostly from #4, but...
-- from #5 whenever I need to do maintenance on the others,
or to experiment with something before trying it on #4:
I keep a TI image of #5 in an almost-virgin state, almost no
tweaks or mods, just w2kpro and winzip, antivirus, firewall,
Mozilla, IE+Googlebar, WebWasher, Spybot, Acrobat.
I restore #5 to this almost-virgin state as needed.
-- other partitions are used for special purposes such as
booting a VPN to the corporate LAN.
In my original post, I'm booting from #5 to do the backup/
restore/defrag on #4.

Acronis/Ilya:
Sorry, but I don't completely understand your response --
but I think it makes no difference, and here's why...
Acronis claims that a TI virtual disk can be used to perform
selective restore of individual files. Well, what I'm doing
is, in effect, a selective restore of individual files (in fact,
of *all* the files). Now, I could understand if someone said
to me, "Certain files have dependencies on each other, so that
selective restores can violate certain internal consistencies."
But that objection doesn't apply, if I'm restoring all of the
files.

The only sense I can make of your answer is this:
when TI performs a conventional full-restore, there's some
"magic data" which is manipulated, in *addition* to the
files and folders (which I'm restoring from the mounted image)
and MFT/directories (which I'm creating by the Win formatting).
SO, here's what I'll try next:
I'll do a conventional full-restore, then delete all the files
and folders, then do a defrag to compact the MFT/dirs,
and then do the selective restore of everything from a mounted
image.

I'll post back in a few hours.

P.S. -- I must point out that, with other vendors' products,
file-based products such as Retrospect, I expect to be able to
format a brand-new partition, add an MBR etc. if needed,
restore all the files on the backup, and have a completely
usable system. Since a TI mounted image is essentially the same
thing as a file-based backup, it still makes no sense to me that
my original plan isn't working. I really need to have a more
detailed explanation.

 

This is really beginning to piss me off.
The bsod is gone, but the explorer problem remains.
Is there something defective in the data provided by the mounted
virtual disk?

 

Hello A-C,

Do you have a SATA disk and >130go ?

I had a problem with a mount virtual drive on a SATA disk (160Go).
I made a test: if I copy the image (Full image) image.tib on a external drive (not SATA and <120Go) and mount the virtual drive and explore it, I can extract the files individually and they are goods. So, the file image.tif is probably correct, but not the mount on a SATA disk or/and a big disk.
I use the build 774. I don't test the build 826.

 

Thanks A-C.

Sounds very similar to my setup. I mainly wanted to verify the "I'm booting from #5 to do the backup/restore/defrag on #4" step. It was not clear from your initial post.

Frankly, I don't see why your approach should not work - unless your are extracted corrupt data from the image?

Have you been moving the image around, from one physical disk to another, from one partition to another, across controllers? If so, please verify that the image has not become corrupted during transfers. Do not rely on the Verify-Image feature in True Image. Calculate an MD5 checksum before and after the copy.

 

As usual, MiniMax is right on target. Something is getting corrupted during the copy or is already corrrupted in the image.

 

MiniMax/armelle
HD is WD800JB, young, ATA.
The image has never been moved: it sits on partition#5, and was
created while booted from #5. To do the copying to #4,
I've tried both msft explorer & Directory Opus.

In case I didn't explicitly say so:
when I do a conventional TI sector-restore, #4 works fine (but
is still fragmented, of course).

At the moment, I'm in the middle of doing a file-name-compare
between the mounted image & the post-file-copy target, to ensure
nothing important was missed in the copying.
If that doesn't show anything wrong, then I'll do a byte-compare
of the virtual disk to the undisturbed TI sector-restore.

 

Hello A-C,

I meant that when you mount the image you are able only to copy the files from it. During the restoration of the partition the necessary changes are made in the Partition Table (if you resize the partition, for example).

Could you please let us know whether the problem appears only with Explorer or with other applications as well?

Thank you.
--
Ilya Toytman

 

Hello armelle,

Thank you for choosing Acronis Disk Backup Software.

First of all, please try the latest build of Acronis True Image. If the problem persists please describe what you are trying to do and what error you get. We will try to help you with the solution.

Thank you.
--
Ilya Toytman

 

Ilya, I'm not changing anything in the partition structure, etc.
As I said, I'm
(a) booting to my "maintenance" partition (partition #5),
(b) imaging my "everyday" partition (partition #4),
(c) re-formatting #4 (i.e., merely making it clean and
unfragmented),
(d) mounting the image,
(e) copying the files from the mounted image to the clean
partition #4 (so I'll have the same files as originally,
but unfragmented and in the order I want),
(f) re-booting to #4.

There should be no need to make changes to the Partition Table.

> only with Explorer or with other applications as well?

Well, when you can't run the windows shell (explore.exe), it's
not easy to run anything else. I can boot and login, then the
taskbar briefly appears, then the desktop is blank.
I can ctrl-alt-del and start Taskman.
In taskman, I can do File==>Run==>Browse, then navigate the
directory structure; but if I do anything which requires
explore.exe, nothing happens (for example, while in the
Open dialog box, try to right click on a folder and select
Open or Explore).
I can do File==>Run==>mmc to reach the services console,
so I can probably run other apps too. I used mmc to run
Event Viewer, and there was nothing unusual in any of the
system Event Records.

And I'm still getting the BSOD at shutdown.

 

Have you thought of doing an "inventory" of partition #4 before you re-format? Something like

Code:

C:> dir C:\ /S > D:\part4.lst
C:> dir C:\ /S /A:D > D:\part4-directories.lst
C:> dir C:\ /S /A:H > D:\part4-hiddenfiles.lst
C:> dir C:\ /S /A:S > D:\part4-systemfiles.lst

After you have re-formatted, and copied everything back to partition #4, redo the dir /s commands, and compare the results (you might need to pipe the results through sort before comparing).

 

Hello A-C,

The problem is that some system files should be stored on soe particular sectors. For example, when you defragment the partition using standard Windows defragmenting procedure you may see the parts marked as "unmovable files". This data should remain on the same place for the proper functioning. It seems that manual defragmentation you describe violate the normal order and consequently leads to problems at bootup.

Thank you.
--
Ilya Toytman

 

Hmm... As I am often heard saying, "I'm no expert, but..." that is totally new to me. Anyone here know what those files would be? I can only think of 2: The Hibernation area, and the paging file. If so, disabling Hibernation and Paging before imaging should work, right?

But - with all respect Ilya - I am not convinced that this is true.

 

Hello MiniMax,

The hibernation and paging files are not included into the image but they are created still. It means that there is an empty hibernation and paging file. There is no use in storing the info from these files in the image but the first bytes of them should be on the particular place. When you use traditional image restoration process Acronis True Image takes care of these details itself but when you do it manually Acronis software cannot control you, of course.

Thank you.
--
Ilya Toytman

 

Ilya, I have to say that these explanations sound bogus.
I'm **NOT** defragging the partition, "manually" or otherwise:
what I'm doing is to gain the *benefits* of defrag (in fact,
better than any real-world defrag) by re-formatting.

There IS no paging or hibernation file involved. I boot from #5
and then delete the paging file on #4 before backup. After I
restore and reboot #4, Windows has no problem in automatically
rebuilding the paging file. That aspect works fine doing both
kinds of restore (copy-files vs. TI-sector restore).

What you're saying is something which violates 15+ years' of
Microsoft Windows experiences and practices world-wide. MSFT
has always told customers how to recover by reformatting and
then restoring from backup -- **WITHOUT** any particular
requirement that the backup/restore be sector-based... in fact,
that's the WHOLE POINT of reformatting, to rebuild all of the
essential structures (MFT, MBR, etc.). If MSFT expected
everyone to use only use sector-based restore, there would be
NO NEED to perform formatting, since a restore of all physical
sectors would automatically contain all the necessary
structural data.

Also, I'm NOT having what you call "problems at bootup". I boot
fine, and can even login and run apps. The problem is that I
can't run the Windows shell.

LET'S CUT TO THE CHASE AND MAKE THIS REALLY SIMPLE...

I will use another vendor's product to do this (backup, then
manual re-format, then file-based restore). If I expend my time
and effort to do this, and it works successfully, will you
THEN agree that there's something wrong with data which I'm
copying from the TI mounted image?

 

Ilya, pardon me for asking, but are you one of the architects
or developers of the product?
If not, could you please escalate this to one of them?
I REALLY need to have a much more authoritative, "bare-metal"
explanation of why this isn't working.

 

A-C,

How do you actually perform the full-file based restore? You should be in an alternative windows install to do a system-restore.

This is how most file-based backups do their restores. Install a "bare" windows in say, WINNT2 or WINDOWS2 directory, then restore all maps from this install, change boot.INI to boot from the original Windows and off you go.

Alternatively you can try a full file restore while booting from another disk, restoring to the temp. disk, then put this temp. disk back as your active disk/partition.

Also keep in mind when copying files back using Acronis explore, that you should copy hidden files/system files, *and* you have to restore security settings from these files as well. A straight windows or Xcopy won't do this.

Furthermore, why use a sector-based backup and then use file based restore (ok I understand you want to optimize...) - Maybe you should look into file-based backups that are out there since years and have working disaster recovery systems based on that. Disaster recovery using Acronis is sector based. Although the image is complete and correct, the file-based "restore" function is probably not designed to do full bare-metal restores.

You might be lucky using the freeware tool XXcopy, using xxcopy [mounted backup drive] [destination disk] /clone /yy - instead of windows explorer copy... of course, like you already stated, you'll need to prepare your disk with MBR.

 

Tonio, please see above, most of your questions are addressed.
Also, I shouldn't have to pay for an additional product (or for
the disk space for an extra set of backups) -- and no extra
product would do what I want -- I've looked.

I always have my Win options set to expose hidden/system files.
And as you'll read above, I already did a file-name compare to
confirm that none were missed. And I've tried two different
utils for copying (explore.exe and Directory Opus).

I'm not sure what you mean when you say, "*and* you have to restore security settings from these files as well."
I'm always Admin on all of the involved partitions.
WHAT OTHER ASPECT OF SECURITY SETTINGS COULD THEN BE MISSED?
RSVP

 

No need to shout.

I was simply stating that you should *copy* the NTFS security information on each file. An administrator of course you have *access* to all files (well not really actually... it's possible to disallow access to admins to certain files - to copy them you'd have to take ownership), but to "clone" a disk, all security information should be in place, or your resulting copy of your system will be compromised, just look at roaming profile directories as an example. Of course, as you are an experienced sysadmin, I don't have to explain you the principles of NTFS security.

I don't quite understand you myself:

If you have another tool that works for you, what is the problem?

and then

You don't have a tool that works for you?

Anyway, you seem to be really agitated at this "problem" of not being able to achieve what you want to do.

I hope you find a way.

 

Hi Tonio - I won't pretend to be able to read A-C's mind, but I would get agitated as well, but primarily about my own inability to understand what it going on...

You sit there at your desk, with many years of computer experience and feel pretty good about yourself. Then along comes this piece of software, where you can pick and choose, copy and paste files from a backup image, onto a clean system partition (not boot! system partition). "Ahh" you think, "here is an interesting method to maintain the ultimate file layout!". And then you discover that it does not work?!?!?

So who do you blame? My first reaction is always to blame myself. Maybe my mental image of this aspects of computer technology is flawed. Maybe I do not fully understand how the boot process, how Windows start-up, which requirements it have on the disk layout and content, etc, etc. So I go back to the drawing board, do some experiments, and I am unable to see any mistakes in my understanding of the technology.

Now, who do I blame next? I blame the only unknown in the scenario: the new piece of software. The worst that can happen then, is that someone tells me that the software is not to blame, that it is my understanding of the technology that is wrong. And to make matters worse, without telling me where I am mistaken! S/he could be perfectly right and I could be wrong, but it leaves me with no way to solve the problem!! And then I get stressed, agitated and outright unreasonable. On the other hand, if s/he is able to tell me where I am wrong, or just point me in a direction that allows me to figure it out myself, then I will be eternal grateful and will forever hold that person in high esteem. If it turns out to be a permission problem, then you Tonio, will become one of those persons

Maybe it is flaw in my personality that I can work myself up over dead things like this, but it is a personality that has served me well, from the day when I first ripped apart a mechanical alarm clock to see how it worked, 'till today where this thirst for knowledge has landed me a job I truely love.

 

Tonio & Minimax

Tonio,
-- I wasn't being irate with you. I upcased that one line cuz i
wanted to emphasize that the line was a question for which I
strongly wanted an answer.
-- My point about being Admin (capital-A) is that I'm the only
user on this machine (although Admin is renamed on the partition
in question). Regardless, I'm not having a problem accessing
files -- as described, the only problems are (a) the shutdown
BSOD, and (b) that explorer.exe refuse to run.
-- Yes, I am an experienced SA, but not in NT; in NT I'd merely
describe myself as a "99-th percentile knowledgable user".
-- When I said "I will use another vendor's product", I meant
that I can use e.g. Retrospect to demonstrate that a format
followed by file-based restore should result in a usable system;
nonetheless, as I said, I don't want my *usual* practice to be
that I must run other backups in addition to TI.
-- When I said "no extra product would do what I want", I meant
that I haven't been able to find a product which will let me
specify a layout/ordering of files to be restored to a "clean"
partition (i.e. formatted or newly defragged).
-- "really agitated"? You bet your ass, I am. I've spent
almost all of the last 80 hours dealing with this, barely sleeping.
And it's not helped by the kind of smoke&mirrors answers I'm
getting from Acronis.
-- see my next post, about a test which implies that TI is fubar.

btw, *your* suggestions at least convey possible specific
avenues of investigation, thank you.

MiniMax...
Thanks for the sympathy. FWIW, I don't get exasperated when the
problem is my lack of knowledge -- depressed, perhaps, but not
exasperated. I *do* get exasperated when I *do* know enough to
know that something *should* work, but it doesn't, and I'm
getting unsatisfactory answers for the problem.
Re "feel pretty good about yourself", this isn't an ego issue
for me. When this problem finally passes, it won't matter to me
if my own cleverness solved it, and my residual feeling will be
grumpiness, because of the sleep I've lost and the pressing
matters which I've let slide.
Like yourself, I'm *not* one of those jump-to-conclusions
blame-everyone-else amateurs. I've already gone to great
lengths to determine if this is my error.
When I first posted this problem, I was quite willing to believe
that the eventual answer would be that I'd overlooked something.
But the more I've pursued it, the more I'm convinced that
there's something about TI which remains to be revealed.
SEE NEXT POST, to Ilya.

 

Re: virtual disk, PROOF? that TI is fubar

Ilya, here's my latest test...

A: booted to part#5, did a full sector restore to part#4, booted
to part#4 and logged-in with no problems.

B: rebooted to part#5, mounted the part#4 image, and performed
"logical" i.e. file-based copy of all files from image,
directly "on top of" the sector-restored part#4, *without*
otherwise modifying part#4 (no format or defrag, no delete
before copy, etc.).
I copied everything except Recycler and System Volume Info.
Again booted/logged-in to part#4 with no problems.

C: re-enacted step B, *EXCEPT* that, before copying files to
part#4, I *deleted* all the files on part#4 (except Recycler &
SysVol). Note that I mean an ordinary vanilla shell-based
permanent "delete", not a re-format.
THIS TIME the previous problems reappeared (BSOD, explorer).

WHY THIS IMPLIES THAT TI IS FUBAR:
TI's doc clearly states that selective restore of individual
files can be performed by copying from a mounted image. I saw
nothing in the doc which says that selective restore won't work
for certain "special" files, or that it doesn't work for files
which were deleted from the target volume. Yet the above test
implies that there are *some* data which are *not* properly
restored by this selective process.
But I don't yet know *exactly* what's going wrong in the
copying. I did a file-name compare of part#4 to the mounted
image, and it appears that all requested files were copied.