SEO

I have an hpux box running 11i that I need to fsck the root partition
(/). I've booted to single-user mode, but cannot run fsck since / is
still mounted. I am unable to umount it.

Is there a way to boot into some type of tools shell where I can run
fsck from?

thanks,
mike

Hi

> I have an hpux box running 11i that I need to fsck the root partition
> (/). I've booted to single-user mode, but cannot run fsck since / is
> still mounted. I am unable to umount it.
>
> Is there a way to boot into some type of tools shell where I can run
> fsck from?

I do not sit in front of a HPUX machine: As far as I remember the
command "reboot" has an option, which says: "do not sync". So you should
go down to single user mode, sync all other filesystems, and then invoke
reboot with this dedicated option.

This will cause, that the / file system is fsck'ed at the next boot.

Florian

If the root file system is VxFS
it does not need to be fsck'ed.


dk

<mjgorr@gmail.com> wrote in message
news:1106844536.437636.203300@c13g2000cwb.googlegr oups.com...
> I have an hpux box running 11i that I need to fsck the root partition
> (/). I've booted to single-user mode, but cannot run fsck since / is
> still mounted. I am unable to umount it.
>
> Is there a way to boot into some type of tools shell where I can run
> fsck from?
>
> thanks,
> mike
>

Hi Dan

> If the root file system is VxFS
> it does not need to be fsck'ed.
????

Please explain why.

Florian

Hi

> I do not sit in front of a HPUX machine: As far as I remember the
> command "reboot" has an option, which says: "do not sync". So you should
> go down to single user mode, sync all other filesystems, and then invoke
> reboot with this dedicated option.
Addition: it is

reboot -n

Florian

"Florian Anwander" <Spam-interessiert-nicht@mnet-online.de> wrote in message
news:35ufl5F4obfrqU1@individual.net...
> Hi Dan
>
> > If the root file system is VxFS
> > it does not need to be fsck'ed.
> ????
>
> Please explain why.
>


FSCK performs consistency checking and
structural repair on file systems that
by virtue of their design can reach an
inconsistent state following a system
crash -- such as UFS, FFS, ext2, etc..

BTW each file system has its own fsck
utility since they all have different
disk layouts. The fsck command is just
a switch that checks the type of the
file system and calls the appropriate
fs specific fsck -- these are usually
named <fs>_fsck, where fs is one of
ufs, ffs, ext2, etc...

VxFS is a transactional file system
that never goes inconsistent, so it
does not need a consistency check,
except in case of media failure.

Transactional file systems are also
described as "journaling file systems",
since they use write-ahead transaction
logs (aka journals) to stay consistent.

Besides VxFS, well-known examples are
Transarc Episode (defunct), SGI XFS,
and IBM JFS.



dk

Hi Dan

>>>If the root file system is VxFS
>>>it does not need to be fsck'ed.
>>????
>>
>>Please explain why.
> [...]
> VxFS is a transactional file system
> that never goes inconsistent, so it
> does not need a consistency check,
> except in case of media failure.
> [...]
Ok so far the sales men talk...

Sorry, I worked four and a half year in the backlevel support for HP
Germany. I cannot count how often I had to recommend a
"/sbin/fs/vxfs/fsck -o full -y " to my customers; and additionally I
could not explain, why there are a lot of VxFS patches, if VxFS is that
perfect.

Sorry, Dan, but VxFS is a filesystem as any other, with that many bugs
as any other filesystem too...

Florian

"Florian Anwander" <Spam-interessiert-nicht@mnet-online.de> wrote in message
news:35v0lvF4povemU1@individual.net...
> Hi Dan
>
> >>>If the root file system is VxFS
> >>>it does not need to be fsck'ed.
> >>????
> >>
> >>Please explain why.
> > [...]
> > VxFS is a transactional file system
> > that never goes inconsistent, so it
> > does not need a consistency check,
> > except in case of media failure.
> > [...]
>
> Ok so far the sales men talk...


I am a super-heavy duty kernel developer
with half a dozen file systems under my
belt. You really don't know who you are
talking to. I was one of the developers
of VxFS, and I managed two of the top
file system development teams in the
industry (Veritas and SGI).


> Sorry, I worked four and a half year in
> the backlevel support for HP Germany. I
> cannot count how often I had to recommend
> a "/sbin/fs/vxfs/fsck -o full -y " to my
> customers;


You worked in support and you don't know
how to fsck a root file system ?!?!?!?!?

"/sbin/fs/vxfs/fsck -o full -y does a very
different kind of work than ufs or ffs fsck.

Not to mention that it is not obvious to
me that the advice you provided to your
customers was relevant.


> and additionally I could not explain, why
> there are a lot of VxFS patches, if VxFS
> is that perfect.


File systems are tightly coupled with
the kernel, and they need to be kept
up to date every time the kernel is
patched. If you subtract the number
of kernel patches from the number of
VxFS patches, you will be left with a
much lower number.

VxFS is not perfect, nor are any other
journaling file systems perfect. That
does not mean that their transaction
mechanisms don't work.


> Sorry, Dan, but VxFS is a filesystem as
> any other,


VxFS is a file system *UNLIKE* any other
file system currently available in the
market place. No other file system that
I am familiar with comes even in design
quality (and I know a lot of file systems).


> with that many bugs as any other
> filesystem too...


Can you provide hard numbers to back up
your statement?



dk

Oh, a goddess spoke! How could I dare to answer from a human point of
view...

Forgive me!


Dan Koren wrote:

> "Florian Anwander" <Spam-interessiert-nicht@mnet-online.de> wrote in message
> news:35v0lvF4povemU1@individual.net...
>
>>Hi Dan
>>
>>
>>>>>If the root file system is VxFS
>>>>>it does not need to be fsck'ed.
>>>>
>>>>????
>>>>
>>>>Please explain why.
>>>
>>>[...]
>>>VxFS is a transactional file system
>>>that never goes inconsistent, so it
>>>does not need a consistency check,
>>>except in case of media failure.
>>>[...]
>>
>>Ok so far the sales men talk...
>
>
>
> I am a super-heavy duty kernel developer
> with half a dozen file systems under my
> belt. You really don't know who you are
> talking to. I was one of the developers
> of VxFS, and I managed two of the top
> file system development teams in the
> industry (Veritas and SGI).
>
>
>
>>Sorry, I worked four and a half year in
>>the backlevel support for HP Germany. I
>>cannot count how often I had to recommend
>>a "/sbin/fs/vxfs/fsck -o full -y " to my
>>customers;
>
>
>
> You worked in support and you don't know
> how to fsck a root file system ?!?!?!?!?
>
> "/sbin/fs/vxfs/fsck -o full -y does a very
> different kind of work than ufs or ffs fsck.
>
> Not to mention that it is not obvious to
> me that the advice you provided to your
> customers was relevant.
>
>
>
>>and additionally I could not explain, why
>>there are a lot of VxFS patches, if VxFS
>>is that perfect.
>
>
>
> File systems are tightly coupled with
> the kernel, and they need to be kept
> up to date every time the kernel is
> patched. If you subtract the number
> of kernel patches from the number of
> VxFS patches, you will be left with a
> much lower number.
>
> VxFS is not perfect, nor are any other
> journaling file systems perfect. That
> does not mean that their transaction
> mechanisms don't work.
>
>
>
>>Sorry, Dan, but VxFS is a filesystem as
>>any other,
>
>
>
> VxFS is a file system *UNLIKE* any other
> file system currently available in the
> market place. No other file system that
> I am familiar with comes even in design
> quality (and I know a lot of file systems).
>
>
>
>>with that many bugs as any other
>>filesystem too...
>
>
>
> Can you provide hard numbers to back up
> your statement?
>
>
>
> dk
>
>

Ont hing you can tyr is

# fsck -n /dev/vg00/lvolxx (probably lovl3)

The -n does a read-only fsck and will not change anything. You can do
it on a live system. Look at the output. If there are no problems,
then you are fine.

Also, the system does an fsck of / if you cycle power. To be on the
safe side, boot to single user, and then cycle power.

There is probably a more straight forward way, maybe booting in lv
maintenance mode, but I cannot think of what it is off the top of my
head.

BTW, I used to do backline HP-UX support, ott, so the guy from Veritas
is free to flame me 8-) 8-)