L2ARC and ZIL on SSD - 4K alignment?

Sebulon said:
@wblock

To my understanding it was -a -s overriding -b, at least in my own experience.

/Sebulon

Is it necessary that we create gnop devices and add them to the pool first, and only then take them offline to replace with real disk as your methods show? Why can't we add disk directly to the pool without creating gnop devices at all?
 
The .nop devices are needed to force 4KB minimum size for I/O units on disks with 4KB sectors that lie about their physical sector size and report 512 byte sectors instead. I'm not sure what you mean by "replace with real disks"? You can create .nop devices directly on top of the real disks if that's what you're asking.
 
einthusan said:
Is it necessary that we create gnop devices and add them to the pool first, and only then take them offline to replace with real disk as your methods show? Why can't we add disk directly to the pool without creating gnop devices at all?

Because if you create the gnop-provider ontop of the gpt-provider, and use that to create (or add to) the pool, the label in zpool "fall off" after a reboot. Don´t ask me why:)

I´ve asked, no one knows. The "phenomenon" can be read about in, e.g. my thread:
Labels "disappear" after zpool import

And I´ve seen at least a dozen more people describing the same problem, both here on the forum and around mailing-lists. So this procedure of not using the actual drives when you create (or add to) the pool is a workaround that I know works.

/Sebulon
 
Sebulon said:
I´ve asked, no one knows. The "phenomenon" can be read about in, e.g. my thread:
Labels "disappear" after zpool import
/Sebulon

FYI I also experienced the same phenomenon up to FreeBSD 9.0-STABLE #1 r235225 where I now get:

Code:
> zpool status
  pool: tank
 state: ONLINE
  scan: scrub repaired 0 in 1h37m with 0 errors on Wed May 16 00:11:55 2012
config:

	NAME            STATE     READ WRITE CKSUM
	tank            ONLINE       0     0     0
	  raidz1-0      ONLINE       0     0     0
	    gpt/zdisk1  ONLINE       0     0     0
	    gpt/zdisk2  ONLINE       0     0     0
	    gpt/zdisk3  ONLINE       0     0     0

This just happened all of a sudden!
 
gkontos said:
FYI I also experienced the same phenomenon up to FreeBSD 9.0-STABLE #1 r235225 where I now get:

Code:
> zpool status
  pool: tank
 state: ONLINE
  scan: scrub repaired 0 in 1h37m with 0 errors on Wed May 16 00:11:55 2012
config:

	NAME            STATE     READ WRITE CKSUM
	tank            ONLINE       0     0     0
	  raidz1-0      ONLINE       0     0     0
	    gpt/zdisk1  ONLINE       0     0     0
	    gpt/zdisk2  ONLINE       0     0     0
	    gpt/zdisk3 ONLINE       0     0     0

This just happened all of a sudden!

Well, does it matter that these labels fall off? My system seems to be working fine and my labels look much worse then that,

Code:
NAME            STATE     READ WRITE CKSUM
	pool1             ONLINE       0     0     0
	  mirror-0      ONLINE       0     0     0
	    ada0p2      ONLINE       0     0     0
	    ada2p2      ONLINE       0     0     0
	  gpt/disk1      ONLINE       0     0     0
 
Export the pool. Then import it using the -d option to point to /dev/gpt. That will force zpool(8) to scan for devices via GPT labels.

# zpool import -d /dev/gpt pool1
 
Back
Top