I wanted to try the new cache feature, so I recently upgraded to 4.3beta6, in this server user shares were disable, I just had a network map for each disk. To use the cache disk I enabled user shares but because the data is separated in each drive each user share is just for one drive.

 

example of how each share is set up:

 

Share name:  TV

Comments: 

Allocation method:  High-water

Split level:  0

Included disk(s):  disk3

Excluded disk(s):  disk1,disk2,disk4,disk5,disk6,disk7,disk8,disk9

Use cache disk:  Yes

Export mode:  Export read/write, hidden

 

Then I mapped a network drive to each share, cache is working fine but every time I save something to a user share all my drives spin up.

 

This is a screenshot after saving a file to a share on disk1 (this one has cache disable), before saving, disk1,disk2,disk5 and disk6 were spun up, all the others were spun down.

 

 

 

 

This didn’t happen with user shares disable, is there any way I can avoid this?

 

Thanks

 

 

try removing the whole "excluded disks" line?

 

 

I added that line to see if the problem was solved, I was getting the same thing without it.

two things:

 

1) try with just one mapping (the one you will use) - this will show if WINDOWS has something to do with the issue

2) have you rebooted unRAID after setting everything up?

 

 

1- It's not windows, disk9 is not mapped, it's not even part of any user share and still spins up

2- Yes

 

2 more things that may help troubleshoot.

 

This does not happen every time, some times i can save a file to a share and disks stay spun down.

 

There was nothing on the cache disk but last night the mover spun up all drives, i cleared the statistics before going to bed and this morning it was like this

 

 

 

end of syslog:

 

May  5 23:17:28 Tower kernel: mdcmd (171): clear

May  6 03:40:01 Tower logger: mover started

May  6 03:41:01 Tower logger: mover finished

May  6 03:55:41 Tower dhcpcd[1634]: sending DHCP_REQUEST for 192.168.0.104 to 192.168.0.254

May  6 03:55:41 Tower dhcpcd[1634]: dhcpIPaddrLeaseTime=86400 in DHCP server response.

May  6 03:55:41 Tower dhcpcd[1634]: DHCP_ACK received from  (192.168.0.254)

May  6 09:01:11 Tower in.telnetd[4590]: connect from 192.168.0.1 (192.168.0.1)

May  6 09:01:19 Tower login[4591]: ROOT LOGIN  on `pts/0' from `192.168.0.1'

 

 

Maybe this is cause by cache and not user shares, I'm going to disable cache for all shares and test for a few days.

One possible reason for many drives to spin up is the existence of some type of server in your network, such as a media server, that does periodic scanning for media changes.  My SageTV scans all of the configured media folders for new or changed videos every 5 minutes.  Normally, this does not spin the drives up, because of the directory caching, so most of the drives will stay spun down for weeks at a time.  But if I move one or more videos to the unRAID server, then most of the drives will spin up and stay active until the moves are finished, then spin down again.  This is because the disk cache in the server has been flushed by the huge files, and the SageTV media scan requires the directories of all of the media drives to be reread.

 

Here's a relevant thread to help with this:  http://lime-technology.com/forum/index.php?topic=1789.  In it, Tom mentions a tunable parameter vfs_cache_pressure that can provide some improvement in retaining the directories cached.  I haven't done enough testing yet to confirm how much help it provides.

 

The above does not very well explain the activity you are seeing, but may spark an idea?  Your screen shows large files being read from Disk 1, 2 and 6, and written to Disk 1.  It also shows something has made 7 reads and writes to all disks, including Disk 9.  As to the Cache drive, are you sure that when you cleared the statistics, that it cleared the reads and writes for the Cache drive?  I suspect it did not, and that is old activity.

 

Disk1 and 6 are always on working with torrents, disk2 is used by the HTPC on my bedroom so its activity last night is normal, I don’t have any media servers, only HTPCs.

 

The cache disk statistics are not reset by the button, that’s normal according to the release notes from Tom.

 

The 7 reads and writes to each disk were made by the cache mover during the night, even thought there was nothing to move, today I clicked the cache move now button, there was only data to move to disk3, but again it made 7 reads and writes to each other disk waking them all, after that I clicked the button again with nothing on the cache disk and it made 7 reads and writes to each disk, I don’t think this should happen.

 

I’m going to disable cache for now.