problem with reiser file system - questions and request for help - Mandriva

This is a discussion on problem with reiser file system - questions and request for help - Mandriva ; I have a machine running Mandriva 2008. It has been working fine for several months. All file systems are reiser. Now there are warnings on boot up saying that several directories are not accessible. They are /var/lib/random-seed and /var/lib/urpmi and ...

+ Reply to Thread
Page 1 of 2 1 2 LastLast
Results 1 to 20 of 39

Thread: problem with reiser file system - questions and request for help

  1. problem with reiser file system - questions and request for help

    I have a machine running Mandriva 2008. It has been working fine for
    several months. All file systems are reiser.

    Now there are warnings on boot up saying that several directories are not
    accessible.

    They are /var/lib/random-seed and /var/lib/urpmi and a couple others.

    LS of /var/lib shows these directories in flashing red. ls-l shows the
    permissions as a series of ????????.

    I cannot access these directories. chmod says it won't access them.

    Any help will be appreciated.

    What causes this? How can I fix it?

  2. Re: problem with reiser file system - questions and request for help

    On Tue, 15 Apr 2008 14:42:20 -0400, n wrote:

    > They are /var/lib/random-seed and /var/lib/urpmi and a couple others.
    > LS of /var/lib shows these directories in flashing red. ls-l shows the
    > permissions as a series of ????????.
    > I cannot access these directories. chmod says it won't access them.


    Most likely, the contents of these directories has been lost.

    The reiserfs file system uses journalization, to protect meta data only,
    not the file contents. I've previously had problems with corrupted file
    systems, using reiserfs, after kernel lockups, or power failures.

    I've turned off hard drive write caching, and haven't had a problem
    with file system corruption, since.

    To turn off the hard drive write caching, edit /etc/sysconfig/harddisks,
    and change the extra parms line to
    EXTRA_PARAMS=" -W 0 "

    Recovering from the problem depends on what, if any backups you have,
    and what other problems fsck.reiserfs finds. Most likely, I expect
    you're going to have to re-install, as it will be difficult to be
    sure you've found all of the problems.

    I've posted about this here, in the past. While turning off the hard
    drive write caching will reduce performance, for a desktop system, I
    don't notice any difference, except when copying large files.

    When the hard drive cache doesn't get written to disk, and the kernel
    thinks it has been, the file system will become corrupt. This can happen
    either due to a power failure, or a kernel lockup forcing a hard reboot.

    If you are running a high usage file server, the best option is to use a
    fully journalized file system, such as ext3, with a ups, and only stable
    software (i.e. not Mandriva).

    For a desktop system, where you want the latest software, turn off
    the hard drive write caching!

    Regards, Dave Hodgins

    --
    Change nomail.afraid.org to ody.ca to reply by email.
    (nomail.afraid.org has been set up specifically for
    use in usenet. Feel free to use it yourself.)

  3. Re: problem with reiser file system - questions and request forhelp

    On Tue, 15 Apr 2008 15:11:38 -0400, David W. Hodgins wrote:

    > I've turned off hard drive write caching, and haven't had a problem with
    > file system corruption, since.


    Had I known about this before I might have avoided losing my 2006 install
    on my laptop. The battery drained out while in suspend and when it came
    back up...no joy...going to stick to ext3 from now on

  4. Re: problem with reiser file system - questions and request forhelp

    On Tue, 15 Apr 2008 15:11:38 -0400, David W. Hodgins wrote:

    > On Tue, 15 Apr 2008 14:42:20 -0400, n wrote:
    >
    >> They are /var/lib/random-seed and /var/lib/urpmi and a couple others.
    >> LS of /var/lib shows these directories in flashing red. ls-l shows the
    >> permissions as a series of ????????.
    >> I cannot access these directories. chmod says it won't access them.


    snip

    > If you are running a high usage file server, the best option is to use a
    > fully journalized file system, such as ext3, with a ups, and only stable
    > software (i.e. not Mandriva).
    >
    > For a desktop system, where you want the latest software, turn off the
    > hard drive write caching!
    >
    > Regards, Dave Hodgins


    Thank you for your kind and very informative reply.

    Out of curiosity, what would you consider to be stable software?

    Thank you again,


  5. Re: problem with reiser file system - questions and request for help

    n wrote:

    > On Tue, 15 Apr 2008 15:11:38 -0400, David W. Hodgins wrote:
    >
    >> On Tue, 15 Apr 2008 14:42:20 -0400, n wrote:
    >>
    >>> They are /var/lib/random-seed and /var/lib/urpmi and a couple others.
    >>> LS of /var/lib shows these directories in flashing red. ls-l shows the
    >>> permissions as a series of ????????.
    >>> I cannot access these directories. chmod says it won't access them.

    >
    > snip
    >
    >> If you are running a high usage file server, the best option is to use a
    >> fully journalized file system, such as ext3, with a ups, and only stable
    >> software (i.e. not Mandriva).
    >>
    >> For a desktop system, where you want the latest software, turn off the
    >> hard drive write caching!
    >>
    >> Regards, Dave Hodgins

    >
    > Thank you for your kind and very informative reply.
    >
    > Out of curiosity, what would you consider to be stable software?


    Not meaning to steal David's airtime, but what is considered "stable"
    depends a lot on what distribution you're using. If you look at Slackware
    for instance, they still consider the 2.6 generation of kernels to be
    "insufficiently tested". Debian is another quite cautious distribution in
    terms of what they consider "stable".

    Mandriva is quite a cutting edge distro, and as such, its components cannot
    really be considered "stable enough" - in the sense of "sufficiently
    tested" for use in production servers that operate in mission-critical
    situations. Therefore, distributions that are typically chosen for such
    server farms are...:
    - Slackware (which unfortunately still only comes in 32-bit version);
    - Debian GNU/Linux;
    - RedHat Enterprise Linux; and
    - CentOS (which is a freely downloadable RedHat).

    Most of the above server environments also deploy /ext3/ as their filesystem
    of choice. It is however advisable for performance reasons that you turn
    on the /ext3/ balanced tree support using /tune2fs./

    /ext3/ has a balanced tree but does not use it by default, so it needs to be
    turned on by the root user. Doing so brings it fairly up to par
    with /reiserfs/ in terms of performance, and that's quite a huge difference
    with the non-balanced tree usage if you've got a lot of disk I/O going on.

    --
    Aragorn
    (registered GNU/Linux user #223157)

  6. Re: problem with reiser file system - questions and request for help

    On Tue, 15 Apr 2008 15:43:01 -0400, mister b wrote:

    > back up...no joy...going to stick to ext3 from now on


    Even with ext3, I'd still disable the hard drive write caching.

    Regards, Dave Hodgins

    --
    Change nomail.afraid.org to ody.ca to reply by email.
    (nomail.afraid.org has been set up specifically for
    use in usenet. Feel free to use it yourself.)

  7. Re: problem with reiser file system - questions and request for help

    On Tue, 15 Apr 2008 16:14:16 -0400, Aragorn wrote:

    > Not meaning to steal David's airtime, but what is considered "stable"


    Heh, heh. Not to worry. As Mandriva is the only distro I have a lot of
    experience with, I'd have been relying on google searches of linux reviews,
    and other comments, previously posted here.

    As you wrote, Mandriva is a cutting edge distro, with all the latest eye candy,
    and new applications. While that's fine for a home desktop system, it would not
    be suitable for a production environment.

    I should add, my experience is limited to Mandriva Free. I haven't looked at
    their corporate release, and have no idea if it sticks to more stable/tested
    versions. I wouldn't be surprised, if the free version is the testing
    environment, for what gets into the corporate version.

    For anyone thinking turning off hard drive write caching will slow down their
    systems too much, I strongly suggest giving it a try. As I've posted before,
    the only time I notice a difference, is when I'm copying a large file, such as
    a dvd iso, from one file system, to another. For the tiny increase in performance,
    for regular day to day usage, it just isn't worth the risk of file system
    corruption.

    Also, welcome back Aragorn!

    Regards, Dave Hodgins

    --
    Change nomail.afraid.org to ody.ca to reply by email.
    (nomail.afraid.org has been set up specifically for
    use in usenet. Feel free to use it yourself.)

  8. Re: problem with reiser file system - questions and request forhelp

    On Tue, 15 Apr 2008 13:42:20 -0500, n wrote:

    > I have a machine running Mandriva 2008. It has been working fine for
    > several months. All file systems are reiser.
    >
    > Now there are warnings on boot up saying that several directories are
    > not accessible.
    >
    > They are /var/lib/random-seed and /var/lib/urpmi and a couple others.
    >
    > LS of /var/lib shows these directories in flashing red. ls-l shows the
    > permissions as a series of ????????.
    >
    > I cannot access these directories. chmod says it won't access them.
    >
    > Any help will be appreciated.
    >
    > What causes this? How can I fix it?



    Try running reiserfsck against the drive, first using --check, then --fix-
    fixable. If that doesn't work, then use --rebuild-tree.

  9. Re: problem with reiser file system - questions and request for help

    David W. Hodgins wrote:

    > On Tue, 15 Apr 2008 16:14:16 -0400, Aragorn
    > wrote:
    >
    >> Not meaning to steal David's airtime, but what is considered "stable"

    >
    > Heh, heh. Not to worry. As Mandriva is the only distro I have a lot of
    > experience with, I'd have been relying on google searches of linux
    > reviews, and other comments, previously posted here.


    Well, I personally have the most experience with Mandrake - yes, that is
    pre-Mandriva - but as you probably have read elsewhere already, I'm also
    in the process of setting up Gentoo - with Xen, which introduces a few
    kernel differences.

    Other than that we use CentOS on our network's servers - not my personal
    favorite, but sufficiently reliable nevertheless - and I've also had some
    minor experience - from the unprivileged user point-of-view - with Debian,
    Slackware, Fedora Core and SuSE.

    > As you wrote, Mandriva is a cutting edge distro, with all the latest eye
    > candy, and new applications. While that's fine for a home desktop system,
    > it would not be suitable for a production environment.


    I agree on that. Companies tend to lean toward what they believe to be
    stable, while consumergrade desktops typically require "the latest and
    greatest".

    > I should add, my experience is limited to Mandriva Free. I haven't looked
    > at their corporate release, and have no idea if it sticks to more
    > stable/tested versions. I wouldn't be surprised, if the free version is
    > the testing environment, for what gets into the corporate version.


    Actually, I have found very little difference between a freely downloadable
    version of Mandr* and a "boxed & shrinkwrapped" retail version. I have
    also purchased and deployed Mandrake 9.0 ProSuite Edition, which was the
    server-/corporation-oriented release of Mandrake 9.0 - it also came with
    the 9.0 PowerPack Edition on a separate DVD included in the box, while the
    ProSuite itself was still on CD-ROMs.

    As far as I could tell, the kernel, /glibc/ and all the /core-utils/ were
    the same as in the PowerPack release. The only difference kernel-wise was
    that an additional kernel was included with PAE support, so that it could
    be used on 686-compatible machines with up to 64 GB of RAM, whereas the
    PowerPack only had the default non-HIGHMEM kernel, a non-HIGHMEM SMP kernel
    and an "Enterprise" kernel, which had HIGHMEM support without PAE, i.e. up
    to 4 GB of RAM.

    Other than that, the difference was in the supplied software. The ProSuite
    basically had all that the PowerPack had but included some server-specific
    non-free software, such as a trial version of IBM's DB2.

    > For anyone thinking turning off hard drive write caching will slow down
    > their systems too much, I strongly suggest giving it a try. As I've
    > posted before, the only time I notice a difference, is when I'm copying a
    > large file, such as a dvd iso, from one file system, to another. For the
    > tiny increase in performance, for regular day to day usage, it just isn't
    > worth the risk of file system corruption.


    Well, my personal advice would be to first and foremost use a UPS on all
    modern computers, even the ones running that Crimosoft OS. The days of
    MS-DOS where a power failure had no bad consequences are long gone, and a
    decent UPS for a home computer shouldn't be too expensive. Laptops are of
    course exempt from this necessity as they have their built-in battery for
    backup.

    Other than that, apart from my barely-ever-used dinosaur-era laptop and this
    regular consumergrade PC here, I've always opted for SCSI drives. My new
    machine - the Xen/Gentoo box - has a hardware RAID 5 with 4 SAS disks.

    As the RAID controller has a 128 MB battery-backed cache, I ran into strange
    error messages when mounting /XFS/ filesystems on that machine. It did not
    happen with every mount, but only after a large amount of filesystems were
    already mounted. A Google search led me to the explanation for that error
    message on an SGI-hosted webpage, which explained the error message to be
    caused by a default mounting option in the newer versions of /XFS/ when
    used on a hardware device that had its own caching.

    The solution was to mount the filesystems with the "nobarrier" mount option,
    which basically tells the kernel not to force-flush the buffers to the
    actual on-disk filesystems itself. So what happens now is that the kernel
    only flushes the buffers "to disk", with "to disk" meaning that it's now in
    the hands of the RAID controller, and that the RAID controller must in turn
    flush the data to the individual disk stripes.

    The bottom line is that in a set-up such as mine, disabling caching
    altogether - I'm not even sure the RAID card's BIOS supports that - would
    be a bad idea.

    Computers are sensitive equipment, and if we are to trust our data to them,
    then we should all decide for ourselves how much that data is worth to us,
    and whether we are willing to get our hands dirty if something goes wrong.
    There is a lot that can be done to avoid file or filesystem damage, and
    while /ext3/ is the most often used filesystem on contemporary GNU/Linux
    machines, /reiserfs/ and /XFS/ are just as reliable.

    /XFS/ is actually older than /ext3/ and was already where /reiserfs/
    (version 3.6) is today back in 1996, when it was first used as the default
    filesystem for IRIX. /XFS/ is very robust and will not easily damage files
    or directories, but something that _does_ happen with it quite often if you
    experience an unclean shutdown is that you lose the latest changes to your
    settings, and often even the not-so-latest changes. This is because of the
    way /XFS/ works - it aggressively caches and only uses last-minute flushes
    - and the journal rollback. However, I have never lost a data file
    with /XFS./

    /reiserfs/ is pretty reliable. I've only once had a problem with it,
    similar to the problems I've mentioned hereabove regarding /XFS,/ but to a
    lesser degree of annoyance.

    /reiser4/ is still not in the mainstream kernel because it's still quite
    experimental. It works in a rather unusual way - using "dancing trees"
    instead of balanced trees - and thus requires a very aggressive caching and
    buffering not dissimilar to (and probably "wilder" than) /XFS./

    On a sidenote, there hasn't been much news regarding /reiser4/ lately, most
    likely due to Hans Reiser having been arrested for the murder of his
    estranged wife Nina - who is/was of Russian origin - for which he is on
    trial at this very moment. I've been following Henry K. Lee's "live blog"
    from the courtroom on the case, and from what it seems, closing arguments
    to the case are to start on April 15th - that's later today for me, and
    tomorrow for US residents.

    Well, now that I'm on that subject, I can't help but wanting to share my
    opinion on the matter, and as we're all GNU/Linux users, I think most of us
    would be interested to hear what happens to Hans Reiser anyway.

    Okay, my opinion is that the whole trial is a joke. First of all, the judge
    currently presiding over that case is not the first judge to have been
    presented with it. The original judge refused to take the matter to court
    on the basis that there is no body - a fact that has yet remained unaltered
    throughout these entire procedings and their background investigations.

    Also, by reading Hans Reiser's testimony and his behavior as testified by
    witnesses and by the blogging reporter following up on this case, it is
    evident that Reiser suffers from Asperger Syndrome, Obsessive-Compulsive
    Disorder - both of which I understand very well since I have those
    conditions as well, and I can thus relate to Reiser's way of thinking - as
    well as a certain degree of paranoia, caused among other things by Nina
    Reiser's former relationship with a KGB agent, and the fact that Reiser was
    ostensibly tailed by quite sloppy undercover police officers, causing him
    to react suspiciously.

    The DA has done his very best at depicting Nina Reiser as a true saint,
    flawless in every way, while sufficient evidence was presented that she was
    not exactly perfect, nor faitful. She cheated on Hans Reiser with his
    former best friend, who is a self-confessed serial killer, but who was
    never even questioned or brought up as a witness, and her computer showed
    signs of her solliciting for sexual encounters with strangers shortly
    before she disappeared.

    On the other hand, being a clear Aspie, Hans Reiser has not exactly made
    himself popular with the judge, who doesn't seem willing to understand how
    the brain of an autistic person works. In short, the whole case is a joke
    based upon ridiculous allegations and the non-discovery of the body of the
    person who was allegedly murdered, and reeks of a DA looking for a
    conviction in order to increase his chances of getting re-elected. Anyway,
    I just had to get that off my chest, and I thought you guys would probably
    like to hear about it. (I apologize for stealing your bandwidth if you
    don't. ;-))

    So, let's now get back to the filesystems... /ext3/ is of course the
    obvious default choice of most distributions - RedHat's, CentOS's and
    Fedora's /anaconda/ installer refuse to create any of the alternatives and
    while they do recognize pre-existing /XFS,/ /JFS/ or /reiserfs/ partitions,
    they do not allow you to install the system on them or mount them anywhere
    into the directory tree - and is a robust and proven filesystem.

    Yet if you want any real performance from it, you'll need to enable the
    balanced tree algorithm, which will then bring it up to par with /reiserfs/
    - /XFS/ outperforms both of them by a large margin. Without the balanced
    tree algorithm, /ext3/ uses an index-sequential approach similar to the
    older /UFS/ in e.g. Solaris.

    Then there's /ext4.../ I personally wouldn't dabble with that yet unless I
    had sufficient experience in how to set it up, because it's a rather
    bizarre filesystem. Basically, it requires you to have two disks -
    although technically it can also be achieved with a single disk, but that
    would be utterly pointless - and to have a partition on one disk store all
    the files, and the partition on the other disk storing all the directories.

    The idea, as you may have guessed, is then to provide for some kind of
    RAID-0-like performance. My advice is that if you want RAID-0 performance
    without owning a hardware RAID controller, stick with any of the regular
    filesystems and just use Linux RAID instead. ;-)

    > Also, welcome back Aragorn!


    Much appreciated, and glad to be back. :-)

    --
    Aragorn
    (registered GNU/Linux user #223157)

  10. Re: problem with reiser file system - questions and request for help

    Aragorn wrote:
    > David W. Hodgins wrote:
    >
    >> On Tue, 15 Apr 2008 16:14:16 -0400, Aragorn
    >> wrote:
    >>
    >>> Not meaning to steal David's airtime, but what is considered "stable"

    >> Heh, heh. Not to worry. As Mandriva is the only distro I have a lot of
    >> experience with, I'd have been relying on google searches of linux
    >> reviews, and other comments, previously posted here.

    >
    > Well, I personally have the most experience with Mandrake - yes, that is
    > pre-Mandriva - but as you probably have read elsewhere already, I'm also
    > in the process of setting up Gentoo - with Xen, which introduces a few
    > kernel differences.
    >
    > Other than that we use CentOS on our network's servers - not my personal
    > favorite, but sufficiently reliable nevertheless - and I've also had some
    > minor experience - from the unprivileged user point-of-view - with Debian,
    > Slackware, Fedora Core and SuSE.
    >
    >> As you wrote, Mandriva is a cutting edge distro, with all the latest eye
    >> candy, and new applications. While that's fine for a home desktop system,
    >> it would not be suitable for a production environment.

    >
    > I agree on that. Companies tend to lean toward what they believe to be
    > stable, while consumergrade desktops typically require "the latest and
    > greatest".
    >
    >> I should add, my experience is limited to Mandriva Free. I haven't looked
    >> at their corporate release, and have no idea if it sticks to more
    >> stable/tested versions. I wouldn't be surprised, if the free version is
    >> the testing environment, for what gets into the corporate version.

    >
    > Actually, I have found very little difference between a freely downloadable
    > version of Mandr* and a "boxed & shrinkwrapped" retail version. I have
    > also purchased and deployed Mandrake 9.0 ProSuite Edition, which was the
    > server-/corporation-oriented release of Mandrake 9.0 - it also came with
    > the 9.0 PowerPack Edition on a separate DVD included in the box, while the
    > ProSuite itself was still on CD-ROMs.
    >
    > As far as I could tell, the kernel, /glibc/ and all the /core-utils/ were
    > the same as in the PowerPack release. The only difference kernel-wise was
    > that an additional kernel was included with PAE support, so that it could
    > be used on 686-compatible machines with up to 64 GB of RAM, whereas the
    > PowerPack only had the default non-HIGHMEM kernel, a non-HIGHMEM SMP kernel
    > and an "Enterprise" kernel, which had HIGHMEM support without PAE, i.e. up
    > to 4 GB of RAM.
    >
    > Other than that, the difference was in the supplied software. The ProSuite
    > basically had all that the PowerPack had but included some server-specific
    > non-free software, such as a trial version of IBM's DB2.
    >
    >> For anyone thinking turning off hard drive write caching will slow down
    >> their systems too much, I strongly suggest giving it a try. As I've
    >> posted before, the only time I notice a difference, is when I'm copying a
    >> large file, such as a dvd iso, from one file system, to another. For the
    >> tiny increase in performance, for regular day to day usage, it just isn't
    >> worth the risk of file system corruption.

    >
    > Well, my personal advice would be to first and foremost use a UPS on all
    > modern computers, even the ones running that Crimosoft OS. The days of
    > MS-DOS where a power failure had no bad consequences are long gone, and a
    > decent UPS for a home computer shouldn't be too expensive. Laptops are of
    > course exempt from this necessity as they have their built-in battery for
    > backup.
    >
    > Other than that, apart from my barely-ever-used dinosaur-era laptop and this
    > regular consumergrade PC here, I've always opted for SCSI drives. My new
    > machine - the Xen/Gentoo box - has a hardware RAID 5 with 4 SAS disks.
    >
    > As the RAID controller has a 128 MB battery-backed cache, I ran into strange
    > error messages when mounting /XFS/ filesystems on that machine. It did not
    > happen with every mount, but only after a large amount of filesystems were
    > already mounted. A Google search led me to the explanation for that error
    > message on an SGI-hosted webpage, which explained the error message to be
    > caused by a default mounting option in the newer versions of /XFS/ when
    > used on a hardware device that had its own caching.
    >
    > The solution was to mount the filesystems with the "nobarrier" mount option,
    > which basically tells the kernel not to force-flush the buffers to the
    > actual on-disk filesystems itself. So what happens now is that the kernel
    > only flushes the buffers "to disk", with "to disk" meaning that it's now in
    > the hands of the RAID controller, and that the RAID controller must in turn
    > flush the data to the individual disk stripes.
    >
    > The bottom line is that in a set-up such as mine, disabling caching
    > altogether - I'm not even sure the RAID card's BIOS supports that - would
    > be a bad idea.
    >
    > Computers are sensitive equipment, and if we are to trust our data to them,
    > then we should all decide for ourselves how much that data is worth to us,
    > and whether we are willing to get our hands dirty if something goes wrong.
    > There is a lot that can be done to avoid file or filesystem damage, and
    > while /ext3/ is the most often used filesystem on contemporary GNU/Linux
    > machines, /reiserfs/ and /XFS/ are just as reliable.
    >
    > /XFS/ is actually older than /ext3/ and was already where /reiserfs/
    > (version 3.6) is today back in 1996, when it was first used as the default
    > filesystem for IRIX. /XFS/ is very robust and will not easily damage files
    > or directories, but something that _does_ happen with it quite often if you
    > experience an unclean shutdown is that you lose the latest changes to your
    > settings, and often even the not-so-latest changes. This is because of the
    > way /XFS/ works - it aggressively caches and only uses last-minute flushes
    > - and the journal rollback. However, I have never lost a data file
    > with /XFS./
    >
    > /reiserfs/ is pretty reliable. I've only once had a problem with it,
    > similar to the problems I've mentioned hereabove regarding /XFS,/ but to a
    > lesser degree of annoyance.
    >
    > /reiser4/ is still not in the mainstream kernel because it's still quite
    > experimental. It works in a rather unusual way - using "dancing trees"
    > instead of balanced trees - and thus requires a very aggressive caching and
    > buffering not dissimilar to (and probably "wilder" than) /XFS./
    >
    > On a sidenote, there hasn't been much news regarding /reiser4/ lately, most
    > likely due to Hans Reiser having been arrested for the murder of his
    > estranged wife Nina - who is/was of Russian origin - for which he is on
    > trial at this very moment. I've been following Henry K. Lee's "live blog"
    > from the courtroom on the case, and from what it seems, closing arguments
    > to the case are to start on April 15th - that's later today for me, and
    > tomorrow for US residents.
    >
    > Well, now that I'm on that subject, I can't help but wanting to share my
    > opinion on the matter, and as we're all GNU/Linux users, I think most of us
    > would be interested to hear what happens to Hans Reiser anyway.
    >
    > Okay, my opinion is that the whole trial is a joke. First of all, the judge
    > currently presiding over that case is not the first judge to have been
    > presented with it. The original judge refused to take the matter to court
    > on the basis that there is no body - a fact that has yet remained unaltered
    > throughout these entire procedings and their background investigations.
    >
    > Also, by reading Hans Reiser's testimony and his behavior as testified by
    > witnesses and by the blogging reporter following up on this case, it is
    > evident that Reiser suffers from Asperger Syndrome, Obsessive-Compulsive
    > Disorder - both of which I understand very well since I have those
    > conditions as well, and I can thus relate to Reiser's way of thinking - as
    > well as a certain degree of paranoia, caused among other things by Nina
    > Reiser's former relationship with a KGB agent, and the fact that Reiser was
    > ostensibly tailed by quite sloppy undercover police officers, causing him
    > to react suspiciously.
    >
    > The DA has done his very best at depicting Nina Reiser as a true saint,
    > flawless in every way, while sufficient evidence was presented that she was
    > not exactly perfect, nor faitful. She cheated on Hans Reiser with his
    > former best friend, who is a self-confessed serial killer, but who was
    > never even questioned or brought up as a witness, and her computer showed
    > signs of her solliciting for sexual encounters with strangers shortly
    > before she disappeared.
    >
    > On the other hand, being a clear Aspie, Hans Reiser has not exactly made
    > himself popular with the judge, who doesn't seem willing to understand how
    > the brain of an autistic person works. In short, the whole case is a joke
    > based upon ridiculous allegations and the non-discovery of the body of the
    > person who was allegedly murdered, and reeks of a DA looking for a
    > conviction in order to increase his chances of getting re-elected. Anyway,
    > I just had to get that off my chest, and I thought you guys would probably
    > like to hear about it. (I apologize for stealing your bandwidth if you
    > don't. ;-))
    >
    > So, let's now get back to the filesystems... /ext3/ is of course the
    > obvious default choice of most distributions - RedHat's, CentOS's and
    > Fedora's /anaconda/ installer refuse to create any of the alternatives and
    > while they do recognize pre-existing /XFS,/ /JFS/ or /reiserfs/ partitions,
    > they do not allow you to install the system on them or mount them anywhere
    > into the directory tree - and is a robust and proven filesystem.
    >
    > Yet if you want any real performance from it, you'll need to enable the
    > balanced tree algorithm, which will then bring it up to par with /reiserfs/
    > - /XFS/ outperforms both of them by a large margin. Without the balanced
    > tree algorithm, /ext3/ uses an index-sequential approach similar to the
    > older /UFS/ in e.g. Solaris.
    >
    > Then there's /ext4.../ I personally wouldn't dabble with that yet unless I
    > had sufficient experience in how to set it up, because it's a rather
    > bizarre filesystem. Basically, it requires you to have two disks -
    > although technically it can also be achieved with a single disk, but that
    > would be utterly pointless - and to have a partition on one disk store all
    > the files, and the partition on the other disk storing all the directories.
    >
    > The idea, as you may have guessed, is then to provide for some kind of
    > RAID-0-like performance. My advice is that if you want RAID-0 performance
    > without owning a hardware RAID controller, stick with any of the regular
    > filesystems and just use Linux RAID instead. ;-)
    >
    >> Also, welcome back Aragorn!

    >
    > Much appreciated, and glad to be back. :-)
    >

    Aragorn, and david.

    I have been using Mandriva since version 9.
    I am just a private person with a 3Ghetz home desktop Intel.

    I have always used ext2, otherwise known as Linux-native.
    I understand that this in not a journalized file system.
    I use it as I understand that it remains in place and 'Shred' will
    actually overwrite it, not leaving copies of itself elsewhere.

    I have experienced lots of power failures where I live, yet Mandriva
    always corrects the damage done, whereby I am unaware of ever losing
    data or having problems on reboot.

    I have have always been satisfied with the speed of ext2.
    Am I left in the dark ages here?
    What advantages would a journalized file system offer a home computer
    user over ext2.
    Thanks for any advice.
    Keep smiling I say.
    Don.

  11. Re: problem with reiser file system - questions and request for help

    On Tue, 15 Apr 2008 21:49:40 -0400, don wrote:

    Please snip the parts of an article, that you are not replying to.

    > I have been using Mandriva since version 9.
    > I am just a private person with a 3Ghetz home desktop Intel.


    I have used other distros, asplinux, slackware, gentoo, but have limited
    experience with them. I've also been using this distro, since Mandrake 9.

    > I have always used ext2, otherwise known as Linux-native.
    > I understand that this in not a journalized file system.
    > I use it as I understand that it remains in place and 'Shred' will
    > actually overwrite it, not leaving copies of itself elsewhere.


    Instead of relying on a file wipe, I use an encrypted file system, for
    all data and most temp files. I'm currently using luks with aes-cbc.
    I keep my browser cache files in memory using /dev/shm.

    > I have experienced lots of power failures where I live, yet Mandriva
    > always corrects the damage done, whereby I am unaware of ever losing
    > data or having problems on reboot.


    You've been lucky! Having a journal helps prevent the loss of data, in
    the event of a power loss. The journal gets updated first, then the
    data. If the power goes in between the two (provided the journal write
    has really gone to disk), the update can be finished, or backed out,
    depending on what was being done. Without the journal, you may have corrupt
    files, and not even know it.

    > What advantages would a journalized file system offer a home computer
    > user over ext2.


    Speed for some situations. Safety of data, provided things have been physically
    written to disk when the file system kernel module thinks they have. Also the
    ability to resize the filesystem, without having to backup/restore the data.

    See http://en.wikipedia.org/wiki/Ext3 and http://en.wikipedia.org/wiki/ReiserFS

    My preference, is reiserfs, as I found ext3 taking too long to fsck, during
    bootup. I haven't really tried other journalized filesystems, such as xfs,
    or jfs, so I can't really comment on them.

    I also find using LVM, useful, as it makes it much easier to add another
    drive, and extend an existing logical volume onto it, again, without having
    to backup/restore existing data.

    Regards, Dave Hodgins

    --
    Change nomail.afraid.org to ody.ca to reply by email.
    (nomail.afraid.org has been set up specifically for
    use in usenet. Feel free to use it yourself.)

  12. Re: problem with reiser file system - questions and request for help

    don wrote:

    > Aragorn, and david.
    >
    > I have been using Mandriva since version 9.
    > I am just a private person with a 3Ghetz home desktop Intel.


    Okay, then you most likely will take after the "latest & greatest" features,
    and then Mandriva is a very good distribution for you. ;-)

    > I have always used ext2, otherwise known as Linux-native.
    > I understand that this in not a journalized file system.


    /ext2/ and /ext3/ are basically the same thing, except that /ext3/ has
    journalling, which /ext2/ does not have. As such, you can easily turn on
    the journalling via /tune2fs/ and thus convert an /ext2/ into /ext3/
    without any problems. It also works the other way around, by the way. ;-)

    > I use it as I understand that it remains in place and 'Shred' will
    > actually overwrite it, not leaving copies of itself elsewhere.


    You are talking about "safe deletion of files", right? Well, all UNIX-style
    filesystems do this.

    > I have experienced lots of power failures where I live, yet Mandriva
    > always corrects the damage done, whereby I am unaware of ever losing
    > data or having problems on reboot.


    As David says, then you've been very lucky. I would definitely recommend a
    simple UPS for your situation, because you may have escaped the big
    filesystem calamities up until now, but one day you won't be able to walk
    away from it so easily anymore.

    > I have have always been satisfied with the speed of ext2.
    > Am I left in the dark ages here?


    Ehm... A little, yes. Performance-wise, there is very little to lose
    from converting your /ext2/ into /ext3./ It can be done without having to
    reformat or reinstall. Just read the /man/ page for /tune2fs./

    And while you're at it, you can experiment with turning on the by default
    disabled balanced tree in /ext2/ or /ext3,/ making your filesystem even
    faster.

    > What advantages would a journalized file system offer a home computer
    > user over ext2.


    If you experience an unclean shutdown with /ext2,/ then it takes a long time
    for /fsck/ to repair the damage, and you may be required to manually
    intervene. You may end up with a lot of obscure files under */lost+found,*
    which you will have to examine one by one in order to know what they
    contain and what directory they used to belong to.

    A journal takes that problem off of your hands under most circumstances, by
    logging the transaction before it occurs. So if the filesystem needs
    repairing, the journal will tell it to what state it should be brought
    back.

    The main difference between the journalling of /ext3/ and that
    of /reiserfs,/ /XFS/ or /JFS/ is that the latter three only keep a journal
    of the metadata, whereas /ext3/ logs both the metadata and the data, making
    it more robust in the long term. I believe David has explained that
    already in an earlier post.

    On this computer here, and on my laptop, I use /reiserfs./ On my Xen/Gentoo
    machine, I use /XFS./ On our network's servers I use /ext3/ - well, I have
    no choice there as CentOS refuses to install on anything else.

    > Thanks for any advice.
    > Keep smiling I say.


    We'll try. :-)

    --
    Aragorn
    (registered GNU/Linux user #223157)

  13. Re: problem with reiser file system - questions and request for help

    Aragorn wrote:

    > don wrote:
    >
    >> I use it as I understand that it remains in place and 'Shred' will
    >> actually overwrite it, not leaving copies of itself elsewhere.

    >
    > You are talking about "safe deletion of files", right? Well, all
    > UNIX-style filesystems do this.


    I feel that my words could be misinterpreted here, so allow me to
    elaborate... ;-)

    Shredding is not the same thing as deleting, as you probably know. However,
    what I was trying to say is that every UNIX-style filesystem - and thus
    this applies to every filesystem upon which you can install GNU/Linux -
    will delete a file without making a backup copy for undeletion.

    On the other hand, deleting a file simply means that all references to the
    file in the filesystem are removed and that the /inode/ becomes available
    again. This does however not mean that by simply using /rm/ or /rmdir,/
    the sectors previously occupied by the file are being zeroed out.

    There are tools that can do that for you - the zeroing out, I mean - but
    this is not the standard filesystem behavior when deleting a file. Whether
    you have /ext2,/ /ext3,/ /ext4,/ /reiserfs,/ /reiser4,/ /XFS/ or /JFS/ is
    irrelevant in this regard.

    Hope this clarifies things... :-)

    --
    Aragorn
    (registered GNU/Linux user #223157)

  14. Re: problem with reiser file system - questions and request for help

    Aragorn wrote:
    > Aragorn wrote:
    >
    >> don wrote:
    >>
    >>> I use it as I understand that it remains in place and 'Shred' will
    >>> actually overwrite it, not leaving copies of itself elsewhere.

    >> You are talking about "safe deletion of files", right? Well, all
    >> UNIX-style filesystems do this.

    >
    > I feel that my words could be misinterpreted here, so allow me to
    > elaborate... ;-)
    >
    > Shredding is not the same thing as deleting, as you probably know. However,
    > what I was trying to say is that every UNIX-style filesystem - and thus
    > this applies to every filesystem upon which you can install GNU/Linux -
    > will delete a file without making a backup copy for undeletion.
    >
    > On the other hand, deleting a file simply means that all references to the
    > file in the filesystem are removed and that the /inode/ becomes available
    > again. This does however not mean that by simply using /rm/ or /rmdir,/
    > the sectors previously occupied by the file are being zeroed out.
    >
    > There are tools that can do that for you - the zeroing out, I mean - but
    > this is not the standard filesystem behavior when deleting a file. Whether
    > you have /ext2,/ /ext3,/ /ext4,/ /reiserfs,/ /reiser4,/ /XFS/ or /JFS/ is
    > irrelevant in this regard.
    >
    > Hope this clarifies things... :-)
    >

    Hello Aragorn and David and mandriva users.
    Such a lot of nice information regarding my inquiry.

    I am not an efficient computer user, I am 81 YO.
    I have young people that also use my computer and they are smart and can
    even read my fingers when I type my root password.

    As we all do, I have private files that I would prefer were overwritten
    rather than deleted. Hence the reason that I use "Shred" from the
    command line, which shows me what it is doing. This will write specific
    and random data over small files a hundred times in seconds. This makes
    me feel rather comfortable and secure.

    Shred does warn that the file to be overwritten must be in place.
    It does not guarantee that it will overwrite the file data in
    journalized file systems as the data may not be where the overwriting is
    taking place. That is the reason that I use ext2.

    It seems an overkill to encrypt these files. Even so if I want to read
    them I need to decrypt and so then they become ordinary files which need
    to be overwritten again.
    I may sound somewhat paranoid, but I do want to understand just how
    secure I am in regard to some smart kid un-deleting my email or any such
    private dialog file. These days they seem to enjoy doing such things.

    I just like to feel comfortable when smart young teenagers are using my
    computer.
    Keep smiling.
    Thanks.
    Don.

  15. Re: problem with reiser file system - questions and request for help

    On Apr 17, 3:06*am, don wrote:
    > Thank you Aragorn. And the group in general.
    > I will take your advice, and I will install anUPS.
    > I have printed out our dialog and I will study what you and David have said.
    > I am beginning to understand the more advanced filing systems.
    > I think now a little better than than the kids do.
    > Although, I never underestimate these young brains, even to the extent
    > of admiring their talent.


    Posted was accurate until he got to the part where the UPS provides
    surge protection. It does not. It also does not claim such
    protection in its numeric specs.

    How does a UPS work? It connects a computer directly to AC mains
    when not in battery backup mode. No effective protection exists in
    that direct connection.

    When voltages drop so low (as to dim incandescent lamps to less than
    50% intensity), then a UPS disconnects from AC mains, takes some time,
    then connects computer to battery. Just battery backup so that 'data
    not saved on disk' can be saved. Data on disk was not at risk.

    BTW, UPS can also identify a household wiring problem. If powering
    other appliances causes that UPS to switch to batteries, then
    household wiring has a problem; probably a loose connection.

    BTW, do not connect motorized appliances to a UPS. That UPS power
    in battery backup mode is so 'dirty' as to even harm small electric
    motors. But same 'dirty' electricity is not harmful to computers
    because computers are so robust. It is called a computer grade UPS
    because its power is safe for computers but too 'dirty' for -
    potentially harmful to - small electric motors.

    UPS is only for data protection. It does not even claim hardware
    protection in numeric specs. That hardware protection is a popular
    urban myth.

  16. Re: problem with reiser file system - questions and request for help

    w_tom wrote:

    > On Apr 17, 3:06*am, don wrote:
    >> Thank you Aragorn. And the group in general.
    >> I will take your advice, and I will install anUPS.
    >> I have printed out our dialog and I will study what you and David have
    >> said. I am beginning to understand the more advanced filing systems.
    >> I think now a little better than than the kids do.
    >> Although, I never underestimate these young brains, even to the extent
    >> of admiring their talent.

    >
    > Posted was accurate until he got to the part where the UPS provides
    > surge protection. It does not. It also does not claim such
    > protection in its numeric specs.


    I beg to differ. I have two UPSes here at home, and they both have surge
    protection. This does of course not mean that they can withstand a
    lightning impact, but it's enough to smoothen out the average voltage
    differences of what you call "dirty" power - we call them brown-outs and
    power spikes.

    > How does a UPS work? It connects a computer directly to AC mains
    > when not in battery backup mode. No effective protection exists in
    > that direct connection.


    Except if you have one that has surge protection, and mine do. ;-) They
    even have power sockets that are not backed up by the battery but that
    still offer surge protection.

    Additionally, the heavier (and far more expensive) brand of UPSes are
    route-through, which means that the power going to the UPS outlet - i.e.
    the connector to which you connect your computer - *always* passes through
    the battery. Such UPSes also require the battery to be loaded to at least
    half their capacity when they are new before you power on the computer.

    > When voltages drop so low (as to dim incandescent lamps to less than
    > 50% intensity), then a UPS disconnects from AC mains, takes some time,
    > then connects computer to battery.


    This timespan needed to switch from AC to battery is short enough not to
    starve the circuits in your computer, though.

    > Just battery backup so that 'data not saved on disk' can be saved. Data
    > on disk was not at risk.


    With most UPS devices, you get a USB or serial UART cable, which you can
    hook up to your computer. You then need a driver or a program talking to
    the UPS, allowing it to signal a power failure and trigger an automated
    shutdown procedure of the computer.

    For the OP's sake: such a serial connection is not required though, but then
    you'll have to take it upon yourself to shut down your computer as quickly
    as possible once the power drops away. Typically, you have about 5 minutes
    for a computer and monitor, and possibly ten minutes or more if the monitor
    is not connected to the UPS (or to a different UPS).

    > [...]
    >
    > UPS is only for data protection. It does not even claim hardware
    > protection in numeric specs. That hardware protection is a popular
    > urban myth.


    I never said it protects your hardware. What I did say - and what I will
    say again - is that many of them - but not all - contain a surge protection
    to even out the brown-outs and the spikes, given that they fall within a
    specified range. If a brown-out drops below a certain voltage, the UPS
    will disconnect from the AC and engage the backup battery.

    If on the other hand there is a spike up to a specified (and sometimes
    user-selectable) voltage, the UPS will trap the surplus voltage and route
    it to ground.

    The above two measures are not for hardware protection but for data
    protection. A power spike might not necessarily damage your computer, but
    it can corrupt your in-memory data, just as cosmic radiation and strong
    electromagnetic fields - e.g. from overhead high voltage cables - can.

    --
    Aragorn
    (registered GNU/Linux user #223157)

  17. Re: problem with reiser file system - questions and request for help

    On Wed, 16 Apr 2008 22:46:32 -0400, don wrote:

    > It seems an overkill to encrypt these files. Even so if I want to read
    > them I need to decrypt and so then they become ordinary files which need
    > to be overwritten again.


    The decrypted versions of the files only exist in ram, and possibly the
    swap, which can also be encrypted.

    I keep most data files on an encrypted filesystem, and only need to
    supply the passphrase, when mounting the filesystem, which is done by
    my ~./bash_profile

    Regards, Dave Hodgins

    --
    Change nomail.afraid.org to ody.ca to reply by email.
    (nomail.afraid.org has been set up specifically for
    use in usenet. Feel free to use it yourself.)

  18. Re: problem with reiser file system - questions and request for help

    on Thursday 17 April 2008 15:43
    in the Usenet newsgroup alt.os.linux.mandriva
    Aragorn wrote:

    > don wrote:
    >
    >> Shred does warn that the file to be overwritten must be in place.
    >> It does not guarantee that it will overwrite the file data in
    >> journalized file systems as the data may not be where the overwriting is
    >> taking place. That is the reason that I use ext2.

    >
    > If you were to use /reiserfs,/ /XFS,/ or /JFS,/ then there would still not
    > be a risk as those filesystems only log the metadata, not the data itself.
    > Or to put it in other words, they keep a journal of the transactions on
    > the file, but not on the actual contents.

    [snip]

    I have a half remembered idea that one of those is normally configured
    to only journal the metadata but can be configured to journal the
    data as well. It is unlikely to be a problem for don.

    Welcome back Aragorn.


    --
    sig goes here...
    Peter D.

  19. Re: problem with reiser file system - questions and request for help

    Peter D. wrote:

    > on Thursday 17 April 2008 15:43
    > in the Usenet newsgroup alt.os.linux.mandriva
    > Aragorn wrote:
    >
    >> don wrote:
    >>
    >>> Shred does warn that the file to be overwritten must be in place.
    >>> It does not guarantee that it will overwrite the file data in
    >>> journalized file systems as the data may not be where the overwriting is
    >>> taking place. That is the reason that I use ext2.

    >>
    >> If you were to use /reiserfs,/ /XFS,/ or /JFS,/ then there would still
    >> not be a risk as those filesystems only log the metadata, not the data
    >> itself. Or to put it in other words, they keep a journal of the
    >> transactions on the file, but not on the actual contents.

    > [snip]
    >
    > I have a half remembered idea that one of those is normally configured
    > to only journal the metadata but can be configured to journal the
    > data as well. It is unlikely to be a problem for don.


    That would be /XFS then. It's an incredibly feature-laden and powerful
    filesystem - I use it myself but even I can't keep track of all the
    possibilities that it offers. ;-)

    > Welcome back Aragorn.


    Thanks Pete! :-)

    --
    Aragorn
    (registered GNU/Linux user #223157)

  20. Re: problem with reiser file system - questions and request forhelp

    On Thu, 17 Apr 2008 15:19:07 -0700, w_tom wrote:

    > Posted was accurate until he got to the part where the UPS provides
    > surge protection. It does not. It also does not claim such protection
    > in its numeric specs.
    >

    Then they should avoid the model that doesn't provide surge protection.
    Most models do these days.

    > How does a UPS work? It connects a computer directly to AC mains
    > when not in battery backup mode. No effective protection exists in that
    > direct connection.
    >

    Not all do this. Good ones provide regulated power power all the time.

    > When voltages drop so low (as to dim incandescent lamps to less than
    > 50% intensity), then a UPS disconnects from AC mains, takes some time,
    > then connects computer to battery. Just battery backup so that 'data
    > not saved on disk' can be saved. Data on disk was not at risk.
    >

    Some work like that. Some don't.

    > BTW, do not connect motorized appliances to a UPS. That UPS power
    > in battery backup mode is so 'dirty' as to even harm small electric
    > motors. But same 'dirty' electricity is not harmful to computers
    > because computers are so robust. It is called a computer grade UPS
    > because its power is safe for computers but too 'dirty' for -
    > potentially harmful to - small electric motors.
    >

    And here all this time I thought it was because the UPS provided a square
    wave voltage. I don't think I've ever heard of it referred to as dirty.
    And even small electric motors draw a lot of current. I was going to plug
    my drill in and try it, but it draws about 350W, and that with the other
    stuff on my ups would overload it. I am certain it wouldn't harm it
    though. No more so than construction people using inverters that runs off
    the vehicle battery to supply AC for power tools. After all, it's the same
    thing as a UPS.

    > UPS is only for data protection. It does not even claim hardware
    > protection in numeric specs. That hardware protection is a popular
    > urban myth.


    I don't know how old the UPS's you use are, but newer models certainly
    provide a constant surge protected voltage. Now APC backups used to
    operate as you describe and may still, but I quit using them 15 or more
    years ago. They, along with Triplite(sp) were always over rated any way.
    And they did not have surge protection.

    --
    Want the ultimate in free OTA SD/HDTV Recorder? http://mythtv.org
    My Tivo Experience http://wesnewell.no-ip.com/tivo.htm
    Tivo HD/S3 compared http://wesnewell.no-ip.com/mythtivo.htm
    AMD cpu help http://wesnewell.no-ip.com/cpu.php

+ Reply to Thread
Page 1 of 2 1 2 LastLast