Bug#592211: mhddfs-related entries in fstab are not processed under dependency-driven boot (insserv)

2010-10-27 Thread Dmitry Astapov
On Wed, Oct 27, 2010 at 8:03 AM, Dmitry E. Oboukhov un...@debian.orgwrote:


 DA It is not a bug of insserv per se. Rather, insserv needs to be notified
 that when mhddfs mounts are present, fuse becomes a pre-requisite for
 mounting them. See bug #41 for similar discussion for another filesystem
 implementation.

 I doubt that *each* fuse filesystem should notify insserv. I think
 that this notification should be placed into fuse-package
 (fuse-utils). And only if a filesystem requires additional services,
 then it should provide such notifies.


What kind of additional services you have in mind, exactly?

I am not a DD, but in my opinion the following makes sense:
If starting up fuse early has no adverse side-effects, then it is the proper
way to go, and this bug should be merged with #41.
Otherwise (assuming there are some side-effects) the following seems to be
the proper course:
1)Fuse should be started early only if there are some fuse-based mounts in
/etc/fstab
2)Fuse-utils could not possibly know how to detect all possible syntaxes for
fuse-based filesystems in /etc/fstab
3)Therefore, it falls on prospective filesystem packages to provide init.d
scripts that would either implicitly mount fstab entries or bump up
dependencies on fuse so that it would be started earlier.

Upon short investigation, I see no obvious problems in just starting fuse
early. However, again, I am not a Debian Developer.

Could you please take this up with maintainer of fuse-utils?



 Does fuse-utils has a filled bug for this subject?


Seems like #41 and #526115 talk about this very same matter.


 --
 ... mpd is off

 . ''`.   Dmitry E. Oboukhov
 : :’  :   email: un...@debian.org jabber://un...@uvw.ru
 `. `~’  GPGKey: 1024D / F8E26537 2006-11-21
  `- 1B23 D4F8 8EC0 D902 0555  E438 AB8C 00CF F8E2 6537

 -BEGIN PGP SIGNATURE-
 Version: GnuPG v1.4.10 (GNU/Linux)

 iEYEAREDAAYFAkzHspwACgkQq4wAz/jiZTdv0gCg2y8aAjJ9MC/YLkL+KZfFHMGA
 pqUAn1+FP6jb1mjNNxoQZNcfHIUPBj7F
 =FMPs
 -END PGP SIGNATURE-




-- 
Dmitry Astapov


Bug#592211: mhddfs-related entries in fstab are not processed under dependency-driven boot (insserv)

2010-10-27 Thread Dmitry E. Oboukhov
DA It is not a bug of insserv per se. Rather, insserv needs to be notified 
that when mhddfs mounts are present, fuse becomes a pre-requisite for mounting 
them. See bug #41 for similar discussion for another filesystem 
implementation.

DA I doubt that *each* fuse filesystem should notify insserv. I think
DA that this notification should be placed into fuse-package
DA (fuse-utils). And only if a filesystem requires additional services,
DA then it should provide such notifies.

DA What kind of additional services you have in mind, exactly?

not only fuse. for example: network, or runned ssh server.

mhddfs depends only on fuse. So I think that fuse should provide mount
posibility in its init scripts. And FSs like sshfs must provide
additional depends.

DA I am not a DD, but in my opinion the following makes sense:
DA If starting up fuse early has no adverse side-effects, then it is the 
proper way to go, and this bug should be merged with #41.

It is a wrong way in mhddfs's init script to init FUSE. I think that
FUSE must be init in FUSE's script.

DA Otherwise (assuming there are some side-effects) the following seems to be 
the proper course:
DA 1)Fuse should be started early only if there are some fuse-based mounts in 
/etc/fstab

Fuse must assume that user can mount FUSE filesystems from fstab (If
he installed fuse-util). I think.

DA 2)Fuse-utils could not possibly know how to detect all possible syntaxes 
for fuse-based filesystems in /etc/fstab

All syntaxes are identical:

filesystem#something /path fuse options number number

any fuse filesystem is mounted as *fuse* type.


DA -BEGIN PGP SIGNATURE-
DA Version: GnuPG v1.4.10 (GNU/Linux)

DA iEYEAREDAAYFAkzHspwACgkQq4wAz/jiZTdv0gCg2y8aAjJ9MC/YLkL+KZfFHMGA
DA pqUAn1+FP6jb1mjNNxoQZNcfHIUPBj7F
DA =FMPs

please, don't quote signatures :)
-- 
... mpd playing: U.D.O. - Shout It Out

. ''`.   Dmitry E. Oboukhov
: :’  :   email: un...@debian.org jabber://un...@uvw.ru
`. `~’  GPGKey: 1024D / F8E26537 2006-11-21
  `- 1B23 D4F8 8EC0 D902 0555  E438 AB8C 00CF F8E2 6537


signature.asc
Description: Digital signature


Bug#592211: mhddfs-related entries in fstab are not processed under dependency-driven boot (insserv)

2010-10-26 Thread Dmitry E. Oboukhov

tags 592211 unreproducible moreinfo
thanks

Hi, Dmitry!

I don't reproduce the problem.

I use mhddfs only from fstab:

home.uvw.ru:[~]$ grep mhddfs /etc/fstab

mhddfs#/mnt/first_1T,/mnt/second_1T,/mnt/fourth/common /share/share fuse 
user,allow_other,default_permissions,exec 0 0
mhddfs#/mnt/third/common,/mnt/hda3,/share/hdd1 /share/common fuse 
user,allow_other,default_permissions,exec 0 0

And it is mounted fine for each booting. So I set unreproducible tag
for this bugreport.

Pleace, show me a configuration which can't mount mhdd filesystem in
booting.


On 11:29 Sun 08 Aug , Dmitry Astapov wrote:
DA Package: mhddfs
DA Version: 0.1.37
DA Severity: important

DA Since recently, insserv (a.k.a. dependency-driven boot) is the
DA preferred way of booting up Debian machines.

DA With default insserv setup, fuse is started after the local and remote
DA filesystems are mounted up, which means that fuse is not available at
DA the time mhddfs-related entries in /etc/fstab are processed.

DA As a result, none of mhddfs filesystems in /etc/fstab are not mounted at
DA the end of boot up.

DA Bug #41 seems to contain a recipe for fixing this (it is related
DA to another fuse-based filsystem, but root cause seems to be the same)

DA -- System Information:
DA Debian Release: squeeze/sid
DA APT prefers unstable
DA APT policy: (500, 'unstable'), (500, 'testing')
DA Architecture: i386 (i686)

DA Kernel: Linux 2.6.30-1-686 (SMP w/2 CPU cores)
DA Locale: LANG=en_US.UTF8, LC_CTYPE=en_US.UTF8 (charmap=UTF-8)
DA Shell: /bin/sh linked to /bin/bash

DA Versions of packages mhddfs depends on:
DA ii  fuse-utils2.8.4-1Filesystem in USErspace 
(utilities
DA ii  libc6 2.11.1-3   Embedded GNU C Library: Shared 
lib
DA ii  libfuse2  2.8.4-1Filesystem in USErspace library

DA mhddfs recommends no packages.

DA mhddfs suggests no packages.

DA -- no debconf information
-- 
... mpd playing: U.D.O. - They Want War

. ''`.   Dmitry E. Oboukhov
: :’  :   email: un...@debian.org jabber://un...@uvw.ru
`. `~’  GPGKey: 1024D / F8E26537 2006-11-21
  `- 1B23 D4F8 8EC0 D902 0555  E438 AB8C 00CF F8E2 6537


signature.asc
Description: Digital signature


Bug#592211: mhddfs-related entries in fstab are not processed under dependency-driven boot (insserv)

2010-10-26 Thread Dmitry Astapov
On Tue, Oct 26, 2010 at 4:41 PM, Dmitry E. Oboukhov un...@debian.orgwrote:

 Hi, Dmitry!

 I don't reproduce the problem.

 I use mhddfs only from fstab:

 home.uvw.ru:[~]$ grep mhddfs /etc/fstab

 mhddfs#/mnt/first_1T,/mnt/second_1T,/mnt/fourth/common /share/share fuse
 user,allow_other,default_permissions,exec 0 0
 mhddfs#/mnt/third/common,/mnt/hda3,/share/hdd1 /share/common fuse
 user,allow_other,default_permissions,exec 0 0

 And it is mounted fine for each booting. So I set unreproducible tag
 for this bugreport.

 Pleace, show me a configuration which can't mount mhdd filesystem in
 booting.


Configuration:
~$ grep mhddfs /etc/fstab
mhddfs#/storage-1Tb,/storage-500Gb /storage fuse allow_other,mlimit=15G 0 2

Insserv version:
$ apt-cache policy insserv
insserv:
  Installed: 1.14.0-2

Maybe your configuration does not use insserv?

Does your configuration include some hint files for insserv in /etc/init.d?
(is dpkg -L mhddfs | grep etc empty or not?)

If not, could you please verify that your installation lists fuse as
prerequisite for any of the moun* scripts like this:
$ grep '^mount[^:]*:' /etc/init.d/.depend.boot | grep fuse

This was producing empty output for me, until I fixed the bug with quick
workaround:
$ cat /etc/init.d/mount-mhddfs
#! /bin/sh
### BEGIN INIT INFO
# Provides:  mount-storage
# Required-Start:$network fuse
# Required-Stop:
# Default-Start: S
# Default-Stop:
# Short-Description: Force mount of mhddfs filesystem on boot
### END INIT INFO
mount /storage

Notice the Required-Start:$network fuse.

HTH, Dmitry.



 On 11:29 Sun 08 Aug , Dmitry Astapov wrote:
 DA Package: mhddfs
 DA Version: 0.1.37
 DA Severity: important

 DA Since recently, insserv (a.k.a. dependency-driven boot) is the
 DA preferred way of booting up Debian machines.

 DA With default insserv setup, fuse is started after the local and remote
 DA filesystems are mounted up, which means that fuse is not available at
 DA the time mhddfs-related entries in /etc/fstab are processed.

 DA As a result, none of mhddfs filesystems in /etc/fstab are not mounted
 at
 DA the end of boot up.

 DA Bug #41 seems to contain a recipe for fixing this (it is related
 DA to another fuse-based filsystem, but root cause seems to be the same)

 DA -- System Information:
 DA Debian Release: squeeze/sid
 DA APT prefers unstable
 DA APT policy: (500, 'unstable'), (500, 'testing')
 DA Architecture: i386 (i686)

 DA Kernel: Linux 2.6.30-1-686 (SMP w/2 CPU cores)
 DA Locale: LANG=en_US.UTF8, LC_CTYPE=en_US.UTF8 (charmap=UTF-8)
 DA Shell: /bin/sh linked to /bin/bash

 DA Versions of packages mhddfs depends on:
 DA ii  fuse-utils2.8.4-1Filesystem in USErspace
 (utilities
 DA ii  libc6 2.11.1-3   Embedded GNU C Library:
 Shared lib
 DA ii  libfuse2  2.8.4-1Filesystem in USErspace
 library

 DA mhddfs recommends no packages.

 DA mhddfs suggests no packages.

 DA -- no debconf information
 --
 ... mpd playing: U.D.O. - They Want War

 . ''`.   Dmitry E. Oboukhov
 : :’  :   email: un...@debian.org jabber://un...@uvw.ru
 `. `~’  GPGKey: 1024D / F8E26537 2006-11-21
  `- 1B23 D4F8 8EC0 D902 0555  E438 AB8C 00CF F8E2 6537

 -BEGIN PGP SIGNATURE-
 Version: GnuPG v1.4.10 (GNU/Linux)

 iEYEAREDAAYFAkzG2msACgkQq4wAz/jiZTdTfQCgzhr8Xp8DQFdcmfT3gprnVfgS
 +bIAn0UeMSGrs7gIa7rfXsH1JVWx6G79
 =MdVJ
 -END PGP SIGNATURE-




-- 
Dmitry Astapov


Bug#592211: mhddfs-related entries in fstab are not processed under dependency-driven boot (insserv)

2010-10-26 Thread Dmitry E. Oboukhov

DA Configuration:
DA ~$ grep mhddfs /etc/fstab
DA mhddfs#/storage-1Tb,/storage-500Gb /storage fuse allow_other,mlimit=15G 0 2

DA Insserv version:
DA $ apt-cache policy insserv
DA insserv:
DA Installed: 1.14.0-2

DA Maybe your configuration does not use insserv?

Yes, Now I hear about insserv for the first time :)
I use lenny/squeeze mix on my server.

May be it is a bug of insserv?

DA Does your configuration include some hint files for insserv in /etc/init.d? 
(is dpkg -L mhddfs | grep etc empty or not?)

empty, of course :)

DA If not, could you please verify that your installation lists fuse as 
prerequisite for any of the moun* scripts like this:
DA $ grep '^mount[^:]*:' /etc/init.d/.depend.boot | grep fuse

/etc/init.d/.depend.boot - file not found :)

-- 
... mpd playing: U.D.O. - Raiders Of Beyond

. ''`.   Dmitry E. Oboukhov
: :’  :   email: un...@debian.org jabber://un...@uvw.ru
`. `~’  GPGKey: 1024D / F8E26537 2006-11-21
  `- 1B23 D4F8 8EC0 D902 0555  E438 AB8C 00CF F8E2 6537


signature.asc
Description: Digital signature


Bug#592211: mhddfs-related entries in fstab are not processed under dependency-driven boot (insserv)

2010-10-26 Thread Dmitry Astapov
On Tue, Oct 26, 2010 at 8:12 PM, Dmitry E. Oboukhov un...@debian.orgwrote:


 DA Configuration:
 DA ~$ grep mhddfs /etc/fstab
 DA mhddfs#/storage-1Tb,/storage-500Gb /storage fuse allow_other,mlimit=15G
 0 2

 DA Insserv version:
 DA $ apt-cache policy insserv
 DA insserv:
 DA Installed: 1.14.0-2

 DA Maybe your configuration does not use insserv?

 Yes, Now I hear about insserv for the first time :)
 I use lenny/squeeze mix on my server.


Well, unless I am much mistaken, all boxes installed with squeeze would have
insserv enabled by default, so it is better to be prepared.



 May be it is a bug of insserv?


It is not a bug of insserv per se. Rather, insserv needs to be notified that
when mhddfs mounts are present, fuse becomes a pre-requisite for mounting
them. See bug #41 for similar discussion for another filesystem
implementation.



 DA Does your configuration include some hint files for insserv in
 /etc/init.d? (is dpkg -L mhddfs | grep etc empty or not?)

 empty, of course :)

 DA If not, could you please verify that your installation lists fuse as
 prerequisite for any of the moun* scripts like this:
 DA $ grep '^mount[^:]*:' /etc/init.d/.depend.boot | grep fuse

 /etc/init.d/.depend.boot - file not found :)

 --
 ... mpd playing: U.D.O. - Raiders Of Beyond

 . ''`.   Dmitry E. Oboukhov
 : :’  :   email: un...@debian.org jabber://un...@uvw.ru
 `. `~’  GPGKey: 1024D / F8E26537 2006-11-21
  `- 1B23 D4F8 8EC0 D902 0555  E438 AB8C 00CF F8E2 6537

 -BEGIN PGP SIGNATURE-
 Version: GnuPG v1.4.10 (GNU/Linux)

 iEUEAREDAAYFAkzHDAcACgkQq4wAz/jiZTc+1ACfcckEHCJoICEXGj5D0MSjfaim
 iIAAl0qZXheAanTV/VL3wnBEXbPbLU4=
 =OGY7
 -END PGP SIGNATURE-




-- 
Dmitry Astapov


Bug#592211: mhddfs-related entries in fstab are not processed under dependency-driven boot (insserv)

2010-10-26 Thread Dmitry E. Oboukhov

DA It is not a bug of insserv per se. Rather, insserv needs to be notified 
that when mhddfs mounts are present, fuse becomes a pre-requisite for mounting 
them. See bug #41 for similar discussion for another filesystem 
implementation.

I doubt that *each* fuse filesystem should notify insserv. I think
that this notification should be placed into fuse-package
(fuse-utils). And only if a filesystem requires additional services,
then it should provide such notifies.

Does fuse-utils has a filled bug for this subject?

-- 
... mpd is off

. ''`.   Dmitry E. Oboukhov
: :’  :   email: un...@debian.org jabber://un...@uvw.ru
`. `~’  GPGKey: 1024D / F8E26537 2006-11-21
  `- 1B23 D4F8 8EC0 D902 0555  E438 AB8C 00CF F8E2 6537


signature.asc
Description: Digital signature


Bug#592211: mhddfs-related entries in fstab are not processed under dependency-driven boot (insserv)

2010-08-08 Thread Dmitry Astapov
Package: mhddfs
Version: 0.1.37
Severity: important

Since recently, insserv (a.k.a. dependency-driven boot) is the
preferred way of booting up Debian machines.

With default insserv setup, fuse is started after the local and remote
filesystems are mounted up, which means that fuse is not available at
the time mhddfs-related entries in /etc/fstab are processed.

As a result, none of mhddfs filesystems in /etc/fstab are not mounted at
the end of boot up.

Bug #41 seems to contain a recipe for fixing this (it is related
to another fuse-based filsystem, but root cause seems to be the same)

-- System Information:
Debian Release: squeeze/sid
  APT prefers unstable
  APT policy: (500, 'unstable'), (500, 'testing')
Architecture: i386 (i686)

Kernel: Linux 2.6.30-1-686 (SMP w/2 CPU cores)
Locale: LANG=en_US.UTF8, LC_CTYPE=en_US.UTF8 (charmap=UTF-8)
Shell: /bin/sh linked to /bin/bash

Versions of packages mhddfs depends on:
ii  fuse-utils2.8.4-1Filesystem in USErspace (utilities
ii  libc6 2.11.1-3   Embedded GNU C Library: Shared lib
ii  libfuse2  2.8.4-1Filesystem in USErspace library

mhddfs recommends no packages.

mhddfs suggests no packages.

-- no debconf information



-- 
To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org