[ANNOUNCE] autofs 5.0.1 release

[ANNOUNCE] autofs 5.0.1 release

[Ian Kent]

Hi all,

There will always be bugs and the line must be drawn at some point and a
release issued and that time is now.

This release is an important milestone for autofs. At last we are at a point
where can expect Linux autofs to function in a multi-vendor environment
without having to maintain separate configurations. This was no easy task
and was made possible by the help and patience of many people, thank you
all.

I particularly want to thank RedHat for making it possible for me to devote
enough time to the development to make it happen, otherwise this release
would have taken much, much longer. Also, I'd like to thank Jeff Moyer for
his seemingly limitless patience, innovative testing and continual code
reviews.

Known issues
============

Quoted strings in the master map are not yet handled.
This will be delayed until the first bug fix release following the initial
release.

autofs
======

The package can be found at:

ftp://ftp.kernel.org/pub/linux/daemons/autofs/v5

It is autofs-5.0.1.tar.[gz|bz2]

No source rpm is there as it can be produced by using:

rpmbuild -ts autofs-5.0.1.tar.gz

and the binary rpm by using:

rpmbuild -tb autofs-5.0.1.tar.gz

See the INSTALL file for information about configure options and
kernel requirements.

Note:
The rc3 tar and patches have been moved to the "old" subdirectory so any
automated build processes that are looking for the tar or patches in the
location above will need to be updated.

Here are the entries from the CHANGELOG which outline the updates:

20/2/2007 autofs-5.0.1
----------------------
- fix typo in Fix typo in var when removing temp directory.
- remove redundant rpath link option.
- ignore "winbind" if it appears in "automount" nsswitch.conf.
- fix another expire regression introduced in the "mitigate manual umount"
  patch.
- correct check for busy offset mounts before offset umount.
- make double quote handing consistent.
- fix handling of trailing white space in wildcard lookup.
- check fqdn of each interface when matching export access list.
- fix race when setting task done.
- correct return status from do_mkdir.
- fix localhost replicated mounts not working.
- add "condrestart" to RedHat init script.
- add "@network" and .domain.name export check.
- fix display map name in mount entry for "-hosts" map.
- update kernel patches.

Ian

Re: [ANNOUNCE] autofs 5.0.1 release

[Lukas Kolbe]

On Di, 2007-02-20 at 16:18 +0900, Ian Kent wrote:
> Hi all,
> 
> There will always be bugs and the line must be drawn at some point and a
> release issued and that time is now.
> 
> This release is an important milestone for autofs. At last we are at a point
> where can expect Linux autofs to function in a multi-vendor environment
> without having to maintain separate configurations. This was no easy task
> and was made possible by the help and patience of many people, thank you
> all.

Cool! 
We just installed it on our mailserver. Sorry for lack of updates on
that thread, I'll get to it later.

The thing we noticed is that autofs5 doesn't seem to support the
"-symlink"-option anymore, is that correct? For local automounted
directories, it now always uses bind mounts.

This is not really a problem, because our stress-test revealed that the
race-condition is not there anymore, so 1000+ simultanous bind-mounts,
umounts and remounts work flawlessly (for stress-testing purposes I set
the automount timeout to 2 seconds), unless you manually and massively
unmount the directories - this leads to a race-condition in mount(8),
because it doesn't seem to lock /etc/mtab correctly. If I leave
automount alone and don't fuss with the mounts, everything works!

Well, there's no real point in this message other than to say *thank
you*, it works!

-- 
Lukas

Re: [ANNOUNCE] autofs 5.0.1 release

[Ian Kent]

On Tue, 2007-02-20 at 21:56 +0100, Lukas Kolbe wrote:
> On Di, 2007-02-20 at 16:18 +0900, Ian Kent wrote:
> > Hi all,
> > 
> > There will always be bugs and the line must be drawn at some point and a
> > release issued and that time is now.
> > 
> > This release is an important milestone for autofs. At last we are at a point
> > where can expect Linux autofs to function in a multi-vendor environment
> > without having to maintain separate configurations. This was no easy task
> > and was made possible by the help and patience of many people, thank you
> > all.
> 
> Cool! 
> We just installed it on our mailserver. Sorry for lack of updates on
> that thread, I'll get to it later.

No problem.

> 
> The thing we noticed is that autofs5 doesn't seem to support the
> "-symlink"-option anymore, is that correct? For local automounted
> directories, it now always uses bind mounts.

I'll check but I must admit that I've been working with the assumption
that we don't need symlinking anymore so if it's broken it is a
consequence of the new development and I'd rather not fix it. 
 
> 
> This is not really a problem, because our stress-test revealed that the
> race-condition is not there anymore, so 1000+ simultanous bind-mounts,
> umounts and remounts work flawlessly (for stress-testing purposes I set
> the automount timeout to 2 seconds), unless you manually and massively
> unmount the directories - this leads to a race-condition in mount(8),
> because it doesn't seem to lock /etc/mtab correctly. If I leave
> automount alone and don't fuss with the mounts, everything works!

Cool, that's great to hear.
I expect, as always, we will have a few new bug reports as new people
try the release out so 5.0.2 is probably where we want to be. That's
always the way though.

Ian

Re: [ANNOUNCE] autofs 5.0.1 release

[Guillaume Rousse]

Ian Kent wrote:
> Hi all,
> 
> There will always be bugs and the line must be drawn at some point and a
> release issued and that time is now.
> 
> This release is an important milestone for autofs. At last we are at a point
> where can expect Linux autofs to function in a multi-vendor environment
> without having to maintain separate configurations. This was no easy task
> and was made possible by the help and patience of many people, thank you
> all.
> 
> I particularly want to thank RedHat for making it possible for me to devote
> enough time to the development to make it happen, otherwise this release
> would have taken much, much longer. Also, I'd like to thank Jeff Moyer for
> his seemingly limitless patience, innovative testing and continual code
> reviews.
Can i expect some news about the patches I submitted ?

Re: [ANNOUNCE] autofs 5.0.1 release

[Ian Kent]

On Wed, 2007-02-21 at 10:12 +0100, Guillaume Rousse wrote:
> Ian Kent wrote:
> > Hi all,
> > 
> > There will always be bugs and the line must be drawn at some point and a
> > release issued and that time is now.
> > 
> > This release is an important milestone for autofs. At last we are at a point
> > where can expect Linux autofs to function in a multi-vendor environment
> > without having to maintain separate configurations. This was no easy task
> > and was made possible by the help and patience of many people, thank you
> > all.
> > 
> > I particularly want to thank RedHat for making it possible for me to devote
> > enough time to the development to make it happen, otherwise this release
> > would have taken much, much longer. Also, I'd like to thank Jeff Moyer for
> > his seemingly limitless patience, innovative testing and continual code
> > reviews.
> Can i expect some news about the patches I submitted ?

Thanks for the reminder.
I've grabbed them again and I'll have a closer look this weekend.

One thing though, I feel quite strongly about renaming the config file
but maybe I misunderstand what your trying to do so let me check it out
first.

Ian

Re: [ANNOUNCE] autofs 5.0.1 release

[Guillaume Rousse]

Ian Kent wrote:
> Thanks for the reminder.
> I've grabbed them again and I'll have a closer look this weekend.
> 
> One thing though, I feel quite strongly about renaming the config file
> but maybe I misunderstand what your trying to do so let me check it out
> first.
I guess you're refering to
autofs-5.0.1-rc1-cleanup-config-files-names.patch here.
Reading it now, it seems I mixed two issues:
- make ldap_auth configuration file researched in autofs configuration
directory, instead of having to set its full name in autofs
configuration file, avoiding yet another subtitution during installation
- and actually renaming autofs to autofs.conf, mainly for consistency
between files

I understand such changes ocurring after a final release may be
considered a bit drastic, that's why I originally submitted quite long
ago. The same applies to the patch renaming configuration options from
DEFAULT_FOO to just FOO: while cleaner, it breaks existing
configurations. If you don't want to merge them now, please consider
them for a 5.1 release then.

Re: [ANNOUNCE] autofs 5.0.1 release

[Ian Kent]

On Tue, 2007-03-06 at 11:41 +0100, Guillaume Rousse wrote:
> Ian Kent wrote:
> > Thanks for the reminder.
> > I've grabbed them again and I'll have a closer look this weekend.
> > 
> > One thing though, I feel quite strongly about renaming the config file
> > but maybe I misunderstand what your trying to do so let me check it out
> > first.
> I guess you're refering to
> autofs-5.0.1-rc1-cleanup-config-files-names.patch here.
> Reading it now, it seems I mixed two issues:
> - make ldap_auth configuration file researched in autofs configuration
> directory, instead of having to set its full name in autofs
> configuration file, avoiding yet another subtitution during installation
> - and actually renaming autofs to autofs.conf, mainly for consistency
> between files

Give me some time, I'm working on this now.

> 
> I understand such changes ocurring after a final release may be
> considered a bit drastic, that's why I originally submitted quite long
> ago. The same applies to the patch renaming configuration options from
> DEFAULT_FOO to just FOO: while cleaner, it breaks existing
> configurations. If you don't want to merge them now, please consider
> them for a 5.1 release then.

It should be fine.
I liked the original names but am changing it to what you recommended.
I added a bit to accept the old names as well. Oops, it just occured to
me, I forgot to update the man pages.

Ian