[vdr] Re: Default video dir

[Rainer Zocholl <UseNet-Posting-Nospam-74308-@zocki.toppoint.de>]

Emil.Naepflein@philosys.de(Emil Naepflein)  16.01.05 06:02


>On Sun, 16 Jan 2005 10:48:43 +1300, Richard Scobie
><r.scobie@clear.net.nz> wrote:

>> LVM then gives you a unified method of combining as many RAIDs as
>> you like into one volume.

>Xes, you are right. But what happens when one of your RAIDs get really
>corrupted because of a multiple disk error, e.g. through power
>failure. Because you have one large filesystem you loose all
>recordings. This is the same as with the standard VDR recording
>distribution. Because of this I have modified VDRs strategy so that
>each managed filesystem is first filled before going to the next one.
>And I migrate all files including meta files of cutted recordings
>always to on partition.

>If the link support of VDR is dropped then the only solution is to
>mount partitions nested into the directory tree. 
>This very bad if a error occurs and it is also very inconvienient 
>if a disk runs full.

>Just think about having a genre directory "Action". You have mounted a
>disk partion to it and it is getting full. If you add a disk you have
>to either add it as "Action1", or you have to mount it elsewhere and
>use symbolic links on the old "Action" volume to get access to the
>recordings within the old structure. Both is not nice.

But you know where your data will be located physically.

>I know that most people would be happy with the much easier handling,
>but as VDR more and and more extents its functionality in direction of
>video server, 

I too would like to be able to record on a removable media
or a NAS mounted with samba.

>dropping this feature reduces data placement flexibility
>considerable. This also limits the possibilities to limit data-loss on
>disk and filesystem failures.


I understand Klaus intervention, that it is crap
that VDR should generate the links for each file from inside to outside.

CYM suggestion that vdr uses exisiting links on directories depending
on catergories or timers seems to much easierer and "transparent"
for vdr and the user.
But VDR can't be allowed to delete the category directory
as this would distroy the link info...

It's a question of viewpoint if it is good that a recording
fails because its disk is full or an other recordings fails.


>IMHO, it would be better to fix the delete algorithm.

I'm still not convinced that the "unexpected delete" would not
have happened if there were no linked files.

AFAIK:
The timer started too early.
The use stopps the timer to be able to delete the superflous recording.
If the timer would not have been stopped, VDR would have give
an extra warning: "recording active, realy delete?"m (or so).
To avoid that warning, the user had stopped the timer before
the delete command.
After the delete command, the timer was reactivated.
That caused vdr to continue writing into the directory,
which is tagged to be deleted...
That does not show any dependency to multiple drives IMHO.

What i don't understand is:
Why were only the files upto 14:00 deleted and not only those,
which exists at 8:00 or all files at 23:45, when the delete 
was executed at 23:45, when the timer stopped.
What happend at 14:00 to mark all .vdr to be deleted?



Rainer---<=====>                         Vertraulich
             //
           //                              
         <=====>--------------ocholl, Kiel, Germany ------------