Mailing List Archive

[...]
> Things to do before release:
> * Adding fuser to datadisk stop

Done in my own latest version I am enclosing ..
I done it in the slave section. I perform the kill only if the directory is
mounted and if I can unmount it
(otherwise you probably kill all the / linked process (including inetd) ;*))

I fixed a bug too ( a "[ test ] && \" line badly used a if then fi was
needed )

Thomas

Am Don, 01 Jun 2000 schrieb Thomas Mangin:
>
>[...]
>> Things to do before release:
>> * Adding fuser to datadisk stop

----------------------------------------
Content-Type: application/octet-stream; name="datadisk"
Content-Transfer-Encoding: quoted-printable
Content-Description:
----------------------------------------

Hi Thomas,

I had a look at your datadisk script, which is a lot better
than the current datadisk script. I have a few questions:

*) The script is for the 0.5.5 release do you want to
port it to 0.5.6 ? (0.5.6 is not yet released, but
simply get the latest source from CVS)

The differences:
*) "drbdsetup xxx SEC" can fail!
*) *_MNTPOINT variables are not longer part of the
configuration files.

*) I am opting for a new variable in the config files,
called [MASTER|SLAVE]_FSCK. The default would be
"fsck -p -y". For journaling FSs this should not
be declared.

-Philipp

> Hi Thomas,
>
> I had a look at your datadisk script, which is a lot better
> than the current datadisk script. I have a few questions:

;*)

>
> *) The script is for the 0.5.5 release do you want to
> port it to 0.5.6 ? (0.5.6 is not yet released, but
> simply get the latest source from CVS)

Ok, I will as soon as I will have a cvs account

> The differences:
> *) "drbdsetup xxx SEC" can fail!

Ok, but I would be pleased to know why ?

> *) *_MNTPOINT variables are not longer part of the
> configuration files.

That must be a big problem I will read it from /etc/fstab like you are
planning to do.

>
> *) I am opting for a new variable in the config files,
> called [MASTER|SLAVE]_FSCK. The default would be
> "fsck -p -y". For journaling FSs this should not
> be declared.

OK.

Thomas

Am Mon, 05 Jun 2000 schrieb Thomas Mangin:

[...]
>> *) "drbdsetup xxx SEC" can fail!
>
>Ok, but I would be pleased to know why ?

It is failing when there is someone who has the device
openend for reading and writing. (This may be a mounted FS or an
user-land-process accessing the device e.g. dd)

>
>> *) *_MNTPOINT variables are not longer part of the
>> configuration files.
>
>That must be a big problem I will read it from /etc/fstab like you are
>planning to do.

I do not think that this is neccessary. If there is the line in
/etc/fstab, mount and unmount are also working when you omit the
mountpoint. (See the current datadisk in CVS)

-Philipp

This message was sent from Geocrawler.com by "Thomas Mangin" <thomas.mangin@example.com>

Hi philip,

I commited the changed. The previous script is
still in the directory under the name datadisk.old
just to be sure that i didn't erase a
functionnality as my version was based on the
stable version

>> "drbdsetup xxx SEC" can fail!
I don't think this is affect the script as I
already check the returned error code, you can
have a look.

>> *) *_MNTPOINT variables are not longer
part of the
>> configuration files.
REMOVED .. I only use the DEVICE

The [MASTER|SLAVE]_FSCK had been added ..

I now need some feedback to fix the possible bugs
..

Thomas

Geocrawler.com - The Knowledge Archive

Am Son, 11 Jun 2000 schrieb Thomas Mangin:

>
>Hi philip,
>
>I commited the changed. The previous script is
>still in the directory under the name datadisk.old
>just to be sure that i didn't erase a
>functionnality as my version was based on the
>stable version
>
>>> "drbdsetup xxx SEC" can fail!
>I don't think this is affect the script as I
>already check the returned error code, you can
>have a look.
>
>>> *) *_MNTPOINT variables are not longer
>part of the
>>> configuration files.
>REMOVED .. I only use the DEVICE
>
>The [MASTER|SLAVE]_FSCK had been added ..
>
>I now need some feedback to fix the possible bugs

I added the RH_5.x-compability code and added the
command to the [MASTER|SLAVE]_FSCK options, because
there is at least one fsck command (the one of ReiserFS)
that is not accessible by the fsck frontend program.
( And I got the feeling that IBM and SGI are also not going
to integrate their fsck commands with the fsck frontend
either)

-Philipp