Hi,

 

i have a problem with the split levels on my unraid server. Currently i have 1 parity + 4 data discs, disk 1-3 are holding a root share 'tvshows' set up as the following:

 

disk1

------ tvshows

------------ showname

-------------------- season 1

-------------------- season 2

------------ showname

....

disk 2

------ tvshows

------------ showname

-------------------- season 1

....

disk 3

------ tvshows

....

 

i like to group all episodes and infos for the 'showname' together, therefor i configured a 'tvshows' user share, highwater, including disk1-3, and set the split level at 2 .. as far as i understand it it should be right. but regardless what i try new files are genereated on the disk with the most freespace. even if i set the splitlevel to 999 with should prevent splitting at all as far as i recall.

 

any idea what could be the problem?

 

thanks in advance...

I have asked Tom via PM to implement a highwater option that doesnt create new folders. I believe when implemented this will solve your issues

i don't get it ... it sould work, shouldn't it? right now for me, this is completely useless since i have to manage the data trough the disk shares which gets anoyin.

i don't get it ... it sould work, shouldn't it? right now for me, this is completely useless since i have to manage the data trough the disk shares which gets anoyin.

You need to set split level to 1.

I have asked Tom via PM to implement a highwater option that doesnt create new folders. I believe when implemented this will solve your issues

 

I think I finally figured out what you mean by this

 

The only use I can see of "don't split any folders" is when you manually set up a structure such as this:

 

disk1/Video/Movies

disk2/Video/Vacations

disk3/Video/Sports

disk4/video/Plays

 

The user share is thus "Video", and you want each disk to just fill up and when full, that's it, no more storage for that category.  You would again need to manually supply another disk.  For example, suppose disk1 fills up.  In order to supply more storage for more Movies you could add disk5 to the mix by going directly to the disk5 share and creating:

 

disk5/Video/Movies

 

Now next time you copy to the Video/Movies directory, system will notice space available on disk5 and start filling it up.

 

Does this about sum it up?

 

 

thanks tom, i set it to 1 and now it works ... don't tell me why, im pretty sure i tried it bevor with no success !? (but i guess pebcak applies here again )

I have asked Tom via PM to implement a highwater option that doesnt create new folders. I believe when implemented this will solve your issues

 

I think I finally figured out what you mean by this

 

The only use I can see of "don't split any folders" is when you manually set up a structure such as this:

 

disk1/Video/Movies

disk2/Video/Vacations

disk3/Video/Sports

disk4/video/Plays

 

The user share is thus "Video", and you want each disk to just fill up and when full, that's it, no more storage for that category.  You would again need to manually supply another disk.  For example, suppose disk1 fills up.  In order to supply more storage for more Movies you could add disk5 to the mix by going directly to the disk5 share and creating:

 

disk5/Video/Movies

 

Now next time you copy to the Video/Movies directory, system will notice space available on disk5 and start filling it up.

 

Does this about sum it up?

 

 

 

Tom i believe you have put it into words perfectly. Theres just something about Highwater splits that i find confusing so I appologise for bad wording.

 

Also this fits with the scenario where you have:

 

//tower/video

 

as a share but cant set split level to 0 as 0 doesnt mean zero anymore.

It's like this;

 

level 1 - tvshows can split across drives

level 2 - showname can split across drives

level 3 - season # can split across drives

 

So, when you used level 2 you will have the showname splitting with different seasons on different drives but the season stuff stays together, correct?

 

Level 1 means the tvshows can split but the showname has to stay on one drive.

 

I like the idea of no directories splitting. You then just manually create a new directory when you need it and it gets used.

 

Peter

 

yeah i get the premise but its not intuitive especially when you have to remember 0 is a special case dependent on what version you are using. (and i keep forgetting)

 

But this thread has focused it for me.... split across the drives using my manually created folders.

 

It all boils down to i dont want stuff split across 15 - 20 drives just because highwater decided that the best way to do it.

 

Its an old discussion and i think this way is an easy solution in the interim until the magic bullet can be found

I have asked Tom via PM to implement a highwater option that doesnt create new folders. I believe when implemented this will solve your issues

 

I think I finally figured out what you mean by this

 

The only use I can see of "don't split any folders" is when you manually set up a structure such as this:

 

disk1/Video/Movies

disk2/Video/Vacations

disk3/Video/Sports

disk4/video/Plays

 

The user share is thus "Video", and you want each disk to just fill up and when full, that's it, no more storage for that category.  You would again need to manually supply another disk.  For example, suppose disk1 fills up.  In order to supply more storage for more Movies you could add disk5 to the mix by going directly to the disk5 share and creating:

 

disk5/Video/Movies

 

Now next time you copy to the Video/Movies directory, system will notice space available on disk5 and start filling it up.

Does this about sum it up?

 

 

split across the drives using my manually created folders.

 

It all boils down to i dont want stuff split across 15 - 20 drives just because highwater decided that the best way to do it.

 

Its an old discussion and i think this way is an easy solution in the interim until the magic bullet can be found

 

 

Yes. This works for me too.

 

yeah i get the premise but its not intuitive especially when you have to remember 0 is a special case dependent on what version you are using. (and i keep forgetting)

 

New in 4.5-beta5 will be special handling of "Split level" with a value of "0".  To get this special handling, you have to enter the number "0" in the share split level setting - leaving the field blank makes it behave like split level is at maximum (this is the current behavior).

 

So, if you set split level explicitly to "0" what happens is this.  When a new object (file or directory) is created, system will create a list of all disks where the parent directory of the object already exists.  It will then pick the disk to use according to the current settings of the include/exclude mask and allocation method.

 

Also new in -beta5 is an additional allocation strategy called "Fill-up".  With this allocation policy system will keep allocating space from the lowest numbered disk in the set of disks eligible for allocation for the share.  When it fills up, system will start allocating from next lowest numbered disk, and so on.

 

To go along with "Fill-up", also added in -beta5 is a per-share "Min free space" setting.  This value sets a "floor" on the amount of free space which must be present for the system to consider a disk for new object allocation.

 

I'm working on a tech note to be published on the website that explains all this better

To go along with "Fill-up", also added in -beta5 is a per-share "Min free space" setting.  This value sets a "floor" on the amount of free space which must be present for the system to consider a disk for new object allocation.

 

Very cool!  This is exactly what I am in need of!  Thank you.

 

New in 4.5-beta5 will be special handling of "Split level" with a value of "0".  To get this special handling, you have to enter the number "0" in the share split level setting - leaving the field blank makes it behave like split level is at maximum (this is the current behavior).

 

So, if you set split level explicitly to "0" what happens is this.  When a new object (file or directory) is created, system will create a list of all disks where the parent directory of the object already exists.  It will then pick the disk to use according to the current settings of the include/exclude mask and allocation method.

 

Also new in -beta5 is an additional allocation strategy called "Fill-up".  With this allocation policy system will keep allocating space from the lowest numbered disk in the set of disks eligible for allocation for the share.  When it fills up, system will start allocating from next lowest numbered disk, and so on.

 

To go along with "Fill-up", also added in -beta5 is a per-share "Min free space" setting.   This value sets a "floor" on the amount of free space which must be present for the system to consider a disk for new object allocation.

 

I'm working on a tech note to be published on the website that explains all this better

 

Very good!  This is what I have been wanting.  I currently do a dance of including and excluding disks so that one gets filled and then I move to the next one.  This works fine but is a little time consuming.  This Fill-up and Min free space sounds great.

 

The only question I have is about how the Min Free Space works.  Is this going to be applied to the user shares or the disk directly?  I should like to set it so that I have about 10GB left on every disk and then it flips over to the next disk in line.

The only question I have is about how the Min Free Space works.  Is this going to be applied to the user shares or the disk directly?  I should like to set it so that I have about 10GB left on every disk and then it flips over to the next disk in line.

 

It is a User Share setting.

yeah i get the premise but its not intuitive especially when you have to remember 0 is a special case dependent on what version you are using. (and i keep forgetting)

 

New in 4.5-beta5 will be special handling of "Split level" with a value of "0".  To get this special handling, you have to enter the number "0" in the share split level setting - leaving the field blank makes it behave like split level is at maximum (this is the current behavior).

 

So, if you set split level explicitly to "0" what happens is this.  When a new object (file or directory) is created, system will create a list of all disks where the parent directory of the object already exists.  It will then pick the disk to use according to the current settings of the include/exclude mask and allocation method.

 

Also new in -beta5 is an additional allocation strategy called "Fill-up".  With this allocation policy system will keep allocating space from the lowest numbered disk in the set of disks eligible for allocation for the share.  When it fills up, system will start allocating from next lowest numbered disk, and so on.

 

To go along with "Fill-up", also added in -beta5 is a per-share "Min free space" setting.   This value sets a "floor" on the amount of free space which must be present for the system to consider a disk for new object allocation.

 

I'm working on a tech note to be published on the website that explains all this better

 

Excellent functionality. I personally will use at least 2 of these immediately.

 

Nice one