I've been noticing this behavior a lot recently.  Both starting and stopping my array has been very slow lately.  During startup, I'll see lots of ReiserFS journal transactions, and during shut down I'll see this kind of stuff:

 

Oct 28 14:53:57 Tower emhttp: _shcmd: shcmd (500): exit status: 1
Oct 28 14:53:57 Tower emhttp: shcmd (501): rmdir /mnt/disk6 >/dev/null 2>&1
Oct 28 14:53:57 Tower emhttp: _shcmd: shcmd (501): exit status: 1
Oct 28 14:53:57 Tower emhttp: shcmd (502): umount /mnt/disk7 >/dev/null 2>&1

 

As you can see in the syslog, that went on for 21 minutes.  During that time I'll see the reads and sometimes writes on all my disks (data and parity) increment.  Parity checks return no sync errors.

 

I recently switched around my hardware, and I've installed a few add-ons.  Here's my current specs:

 

Hardware: Biostar A760G M2+, AMD Sempron 140, 2 GB Kingston DDR2 800 RAM, Antec Nine Hundred Two case, 9 x single hot swap bays, Corsair 400W PSU, 8-9 Drives (currently 8 drives installed, no cache drive), 2 port PCIe x1 card (both ports used), 4 port Promise TX4 card (one port used for cache drive, when installed).

 

I also just plugged in a new 2 TB WD EARS (jumpered) to start preclearing.  That was after I captured this syslog.

 

Software: unRAID 4.5.6, unMenu (which still only works with the IP address, not 'tower'), cache_dirs, APCUPSD (but no UPS connected...I need to get around to that soon), midnight commander, monthly parity check, clean powerdown.  All except for cache_dirs were installed via unMenu.

 

 

 

This line at the end of that 21 minutes makes me think that cache_dirs is at fault here:

 

Oct 28 15:12:27 Tower cache_dirs: Suspending cache_dirs for 120 seconds to allow for clean shutdown of array
Oct 28 15:12:27 Tower cache_dirs: While suspended, pressing "Stop" on the unRAID management web-interface will shutdown the array

 

Syslog attached.

21mins_syslog_10-28-2010.zip

RAID will not stop the array unless the disks are not "busy"

 

If the array is showing "writes" then it is in use.  You need to FIRST stop the add-on processes, and then stop the array.

 

cache_dirs will pause itself when it see the messages in the system log.  While it is in the middle of a "find" of the contents of your disks, your disks are in use.

 

The other messages you are seeing are normal.  unRAID will attempt to un-mount the disks again and again, until they are idle and the un-mount succeeds.

 

Set cache_dirs to scan less depth, or exclude directories not needed, or add more memory so it all fits in memory and the disks do not get accessed.  All of those might help keep cache_dirs "find" commands from keeping your disks busy.

 

Joe L.

Great, thanks Joe.  Looks like a good excuse to add more RAM to my server

maybe if you know shell scripting you could catch a shut down and force a cache_dirs -q in order to stop cache_dirs before the shutdown.

I cant help you since i know very little shell dcripting but maybe someone else can....

Should I be worried about the journal transactions during start up?  Don't those usually indicate an unclean shut down or some other problem?

maybe if you know shell scripting you could catch a shut down and force a cache_dirs -q in order to stop cache_dirs before the shutdown.

I cant help you since i know very little shell dcripting but maybe someone else can....

That will not help if cache_dirs is in the middle of a "find" command since it is only looks for the flag to quit between "find" commands.   It already looks for the message in the syslog for the server trying to un-mount the disks, so one is no better off.  (And, as noted, suspends itself so the array can stop)

 

Joe L.