I've searched the forums and didn't find the answer to my question.

 

4x1TB disks. 3 as disk, 1 as parity.

 

All format and parity checks were complete earlier last night. I had begun copying data during parity check which didn't look to be a problem

 

 

All shares set to High-Water mark, export read/write. No split level sans 1 dir and that is not having any data being written to it at this time.

 

Copying over 1TB+ of data to different user shares.

 

When looking at MAIN I see the following:

 

Disk 1: Free 400G

Disk 2: Free 1TB

Disk 3: Free 1TB

 

Disk 2/3 are blinking indicating there are active and spun down.

 

Now according to High-Water mark value the data SHOULD stop copying from Disk1 and start moving to Disk 2 when the threshold reaches less than 50%. As it's approach 40% I'm wondering if there isn't something fishy going on.

 

I know I'm 'safe' with the parity disk if the other disk fails. I'm trying to understand what it is doing and verify everything is working as normal.

 

Any idea if this is normal, acceptable behavior or am I just being paranoid?

 

 

Ok, here's what my current theory is on this after a few hours of troubleshooting (and a bit of Jack Daniels).

 

The shares I created were numerous (about 6 of them). I created this off the web page.

 

Now when I started copying data to this none of the shares themselves reached the high-water mark (50% or 450GB), hence why it was all staying on one disk instead of spanning disks.

 

I deleted all that data (dumb on retrospect but I want to test rsync) and redid the share to be one share: data

 

from under data I will create separate directories manually and copy the data there, hence the dir will reach the magic 50% mark and span the disks like they should.

 

In retrospect I could've created a new share called data and moved all the files into it and it would've spanned as a test. Took most of the night but I'll see how performance is with rsync now.

 

I'll give it a whirl and see what happens.

 

not liking rsync copying from ubuntu to unraid.

 

 

I had to mount this as a SMB share to get it to read. Couldn't rsync TO the server directly, had to mount it under /mnt on my ubuntu workstation and then was able to copy stuff to it.

 

 

I'm getting 9MB/sec copy speed to it. Now this is a 10/100 network and I'm showing FD using mii-tool that I've got 100 on both end. Is it supposed to go that slow?

So 8MB is common. This translates into 80mb on my 10/100mb network....man I need to get back to my Cisco classes

8 MB/sec = 64 Mbits/sec... which is doing good on 100Mbps Ethernet.

 

See this post:

 

  http://lime-technology.com/forum/index.php?topic=787.msg21068#msg21068

ok, now I have a problem.

 

I've crossed the high-water mark at 50% of my total:

 

976,762,552  /2 =488GB roughly.

 

I'm at 480GB free now and it is still using only 1 of the disks and has not moved over to one of the other 2.

 

Do I have to IMPLICITLY tell the user-share to take all 3 disks? It was my understanding it would know when it reaches the high-water mark and then activate the other disks in the array and begin to copy data to them.

 

I've been copying non-stop for about 12hrs and I don't want to do this again.

 

Any ideas?

 

i added the 3 disks to the user-share:

 

Included disks: disk1,disk2,disk3

 

do i have to start/stop the array or copy process for it to merge over?

 

Would creating the directories on the disks help?

removed included drives from settings.

 

Added split levels.

 

It's now moving data between disks appropriately.

 

phew.

 

 

Now to my understanding split level was for keeping data together all on one drive, that's a given. What was most curious is why it doesn't branch out once it shows that the high-water mark is reached on the main disk(disk 1) and then subsequently begins filling the other disks as needed.

 

to me I thought it would work like this:

 

1. Create share with all 3 disks

2. Copy data

3. Disk 1 begins taking data

4. Once disk 1 reaches 50% of maximum capacity disk 2 is activated

5. Data continues to copy now to disk 2 until 50% then activate disk 3.

6. Repeat 3-5

 

 

Was I wrong in thinking this?

I would set the included disks (e.g. disk1-3) and split level=999. That works for me always. As far as I know if you don't set split level explicit it will be 0 and everyting will stay on the first disk. If you start using split level the expansion will start.

 

On my second unRAID machine I forgot split level and I could see the first disk filling up. I had to set split level (even during the copy process) and the next disk was in use nearly immediately.

 

Regards

Harald

 

 

I'm at 480GB free now and it is still using only 1 of the disks and has not moved over to one of the other 2.

 

Do I have to IMPLICITLY tell the user-share to take all 3 disks? It was my understanding it would know when it reaches the high-water mark and then activate the other disks in the array and begin to copy data to them.

Many have reported that user shares are broken in the 4.4 final and 4.5beta versions of unRAID.  

 

No, you should not have to IMPLICITLY tell user-shares anything.

 

I'm guessing Tom accidentally broke user-share allocation when he created 4.4final.  (He tried to add one-more feature, without it being in a "beta" version first, and it broke other features of allocation.)

 

His new feature is looking for a "string" in the split level field, and if that  "string" exists in the directory path it acts differently.  I'm guessing whatever you have in "split level" is being interpreted as "keep everything on the same drive" even if that is not your intent.

 

Tom mentioned in the release notes he would be writing a Wiki article describing the new behavior, but as far as I know, he has not yet  posted it.

 

For now, you might want to consider the user share allocation broken... Instead, copy to the "disk" shares... at least you will be able to control the transfers to go where space is really available.

 

Joe L.

Harald: I set it to be split level 2 and it started breaking the data to the other disks. 999 will make it to do what?

 

Joe: What I'm curious about is that will unraid resync the data across the array being 'smart' about where stuff goes. i.e. say "Ok movies are here, but this drive is 1/2 full lets move this here and then games there to this disk instead of this one". Or should I just consider the whole raid one big disk and move stuff around there manually from the terminal window?

 

Harald: I set it to be split level 2 and it started breaking the data to the other disks. 999 will make it to do what?

 

Joe: What I'm curious about is that will unraid resync the data across the array being 'smart' about where stuff goes. i.e. say "Ok movies are here, but this drive is 1/2 full lets move this here and then games there to this disk instead of this one". Or should I just consider the whole raid one big disk and move stuff around there manually from the terminal window?

 

unRAID never moves a file or directory once it has put it somewhere.  The only way to do what you are describing is for you to specify the "disk" shares when copying files to the server.  The "user" shares are great for providing a consolidated view on the lan for playing media, etc.

 

I have my "disk" shares as read/write, but hidden.  I can open up a window on my PC and type: \\tower\disk3 and get to the "Movies," TV,", "MP3" amd "data" folders under it.  I never see the "disk" shares on the LAN.

 

I usually leave the "User" shares as read-only.

 

Joe L.

Harald: I set it to be split level 2 and it started breaking the data to the other disks. 999 will make it to do what?

 

Split level 1 and up is for keeping files together that share that same folder level. With 999 I tell I don't care about files being stored together but please fill up all the disks.

 

Here's a quote from the manual:

 

If you set the Split level to 0 for a share, then all directories/files created under that share will be on the same disk where the share was originally created.

If you set the Split level high, e.g., 999 for a share, then every directory/file created under that share will get placed on a disk according to Allocation method.

 

Regards

Harald

 

 

At some point in time, I seem to remember a split level of 0 was changed to be equivalent to 999.

so 999 is basically scattering it across all disks, just filling up all available space first?

At some point in time, I seem to remember a split level of 0 was changed to be equivalent to 999.

 

this was my understanding as well. Apparently my assumption was wrong and I had to put this in place. I'm more about the storage of data and making sure all disks are not being overloaded than dir structure. My unRAID is just a storage area for files, not really doing too much with it, just need room to grow.

I have my "disk" shares as read/write, but hidden.  I can open up a window on my PC and type: \\tower\disk3 and get to the "Movies," TV,", "MP3" amd "data" folders under it.   I never see the "disk" shares on the LAN.

 

I usually leave the "User" shares as read-only.

 

may I ask you why you do this? IMHO you have to guess on what disk is what to place new files there.

 

Consider a new movie and consider a lot of disks (unRAID supports up to 15 data disks). How do you decide on what disk you want to store this movie? You open up disk1 (argh, this one was full), then you open up disk2 (argh, the music disk), ... you get the idea.

 

Hmm, what's the reason behind this?

 

Thanks

Harald

 

 

If you choose the high water method of allocation then 999 will put all files on one hard drive until it is 50% full then it will place files on the next disk.

I have my "disk" shares as read/write, but hidden.  I can open up a window on my PC and type: \\tower\disk3 and get to the "Movies," TV,", "MP3" amd "data" folders under it.   I never see the "disk" shares on the LAN.

 

I usually leave the "User" shares as read-only.

 

may I ask you why you do this? IMHO you have to guess on what disk is what to place new files there.

 

Consider a new movie and consider a lot of disks (unRAID supports up to 15 data disks). How do you decide on what disk you want to store this movie? You open up disk1 (argh, this one was full), then you open up disk2 (argh, the music disk), ... you get the idea.

 

Hmm, what's the reason behind this?

 

Thanks

Harald

 

 

Actually, I just open up the unRAID web-interface to see which disk has theavailable free space.  At this time, most of my disks are pretty full, but I just last night added a 750Gig drive as disk8.

I can group all the Movies if I desire, but in reality, they are on most of my physical disks.  The "data" folders are only on a few.  I use those for windows-backups, etc.

 

My reasoning behind using actual disk names when storing files is simple: I'm in control.

My reasoning for using "User-Shares" when browsing my server for something to play, simple...Usability... I don't need to remember which physical disk anything is on, and the media players only have one folder to browse.

 

To make it easier on me and my wife, I also run the following script every once in a while.  It makes a "link" to all the "New" movie ISO files on the server to the same name in a "Movies_New" share.  It also links all the movies to various folders each containing a subset-alpha range of titles.  If I'm looking for "WHITE_CHRISTMAS" I can browse to it quickly by choosing the "Movies T-Z" folder.  If I'm just looking to find a movie to watch at random, I can open up the "Movies" folder.  If I want to find something I've recently added to the server, it is in "Movies_New"

 

[pre]

root@Tower:/boot# cat link_movies.sh

# Link in the 50 most recent movies

ls -1tr /mnt/user/Movies | tail -50 | while read a

do

  fil=`ls /mnt/disk*/Movies/$a`

  dir=`dirname $fil`

  if [ ! -d "${dir}_New" ]

  then

      mkdir "${dir}_New"

  fi

  if [ ! -f "${dir}_New/$a" ]

  then

      ln $fil ${dir}_New/

  fi

done

for d in /mnt/disk[0-9]*

do

  cd $d

  rm Movies\ A-E/*.ISO

  rm Movies\ F-J/*.ISO

  rm Movies\ K-O/*.ISO

  rm Movies\ P-S/*.ISO

  rm Movies\ T-Z/*.ISO

  rmdir Movies\ A-E

  rmdir Movies\ F-J

  rmdir Movies\ K-O

  rmdir Movies\ P-S

  rmdir Movies\ T-Z

  mkdir Movies\ A-E Movies\ F-J Movies\ K-O Movies\ P-S Movies\ T-Z

  for i in Movies/[0-9A-E]*.ISO;do ln "$i" Movies\ A-E/; done

  for i in Movies/[F-J]*.ISO;do ln "$i" Movies\ F-J/; done

  for i in Movies/[K-O]*.ISO;do ln "$i" Movies\ K-O/; done

  for i in Movies/[P-S]*.ISO;do ln "$i" Movies\ P-S/; done

  for i in Movies/[T-Z]*.ISO;do ln "$i" Movies\ T-Z/; done

done

[/pre]

Thanks Joe,

 

I was only wondering.

 

On my system all is hidden behind those user-shares. For me this is one of the top-arguments for unRAID (beside incremental storage and the possibility to mix drive types). I use user-shares, and users (some have read-only access, some have r/w accress) some user-shares are bound to a single disk and some spread a bunch of disks. In normal operation I never access single-disks (ok, ok, I just put one disk from my 1st server into the 2nd server because I thought cp would be faster then - this was an error).

 

Regards

Harald