zfs discuss - Nov 2010 - ZFS doesn''t notice errors in mirrored log device?

Hello!

I''ve got a Solaris 10 10/08 Sparc system and use ZFS pool version 15.
I''m
playing
around a bit to make it break.

I''ve created a mirrored "Test" pool using mirrored log
devices:

# zpool create Test \
    mirror /dev/zvol/dsk/data/DiskNr1 /dev/zvol/dsk/data/DiskNr2 \
    log mirror /dev/zvol/dsk/data/LogMirrorDev1
/dev/zvol/dsk/data/LogMirrorDev2

Now I write some garbage to one of the log mirror devices.

# dd if=/dev/urandom of=/dev/zvol/dsk/data/LogMirrorDev2 bs=1024
count=100000

Now I''m checking the status:

# zpool status -v Test
  pool: Test
 state: ONLINE
 scrub: none requested
config:

        NAME                                  STATE     READ WRITE CKSUM
        Test                                  ONLINE       0     0     0
          mirror                              ONLINE       0     0     0
            /dev/zvol/dsk/data/DiskNr1        ONLINE       0     0     0
            /dev/zvol/dsk/data/DiskNr2        ONLINE       0     0     0
        logs
          mirror                              ONLINE       0     0     0
            /dev/zvol/dsk/data/LogMirrorDev1  ONLINE       0     0     0
            /dev/zvol/dsk/data/LogMirrorDev2  ONLINE       0     0     0

errors: No known data errors

No error? Because th in-memory data is still correct? Let''s export and
import it, which should clear this, right?

# zpool export Test
# zpool import -d /dev/zvol/dsk/data -a

# zpool status -v Test
  pool: Test
 state: ONLINE
 scrub: none requested
config:

        NAME                                  STATE     READ WRITE CKSUM
        Test                                  ONLINE       0     0     0
          mirror                              ONLINE       0     0     0
            /dev/zvol/dsk/data/DiskNr1        ONLINE       0     0     0
            /dev/zvol/dsk/data/DiskNr2        ONLINE       0     0     0
        logs
          mirror                              ONLINE       0     0     0
            /dev/zvol/dsk/data/LogMirrorDev1  ONLINE       0     0     0
            /dev/zvol/dsk/data/LogMirrorDev2  ONLINE       0     0     0

errors: No known data errors

Still no error? Do I need to scrub first?

# sudo zpool scrub Test ; sleep 5 ; zpool status -v Test
  pool: Test
 state: ONLINE
 scrub: scrub completed after 0h0m with 0 errors on Fri Nov 12 15:20:26 2010
config:

        NAME                                  STATE     READ WRITE CKSUM
        Test                                  ONLINE       0     0     0
          mirror                              ONLINE       0     0     0
            /dev/zvol/dsk/data/DiskNr1        ONLINE       0     0     0
            /dev/zvol/dsk/data/DiskNr2        ONLINE       0     0     0
        logs
          mirror                              ONLINE       0     0     0
            /dev/zvol/dsk/data/LogMirrorDev1  ONLINE       0     0     0
            /dev/zvol/dsk/data/LogMirrorDev2  ONLINE       0     0     0

errors: No known data errors

Hm. Why are there no errors shown for the logs devices?

Thanks a lot,

Alexander
-- 
?    Lifestream (Twitter, Blog, ?) ? http://alexs77.soup.io/     ?
? Chat (Jabber/Google Talk) ? a.skwar at gmail.com , AIM: alexws77  ?
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://mail.opensolaris.org/pipermail/zfs-discuss/attachments/20101112/a54c0ea3/attachment.html>

On Nov 12, 2010, at 5:21 PM, Alexander Skwar wrote:
> Hm. Why are there no errors shown for the logs devices?
You need to crash you machine while log devices are in use, then you''ll
see some reads on the next reboot. "In use" here means that system is
actively writing to log devices at the time of crash. And as long as log blocks
are short-lived, it may be a challenge to overwrite those that are still in use
at the time of crash, unless you do it after a crash and before next boot..

Victor

> From: zfs-discuss-bounces at opensolaris.org [mailto:zfs-discuss-
> bounces at opensolaris.org] On Behalf Of Alexander Skwar
> 
> I''ve got a Solaris 10 10/08 Sparc system and use ZFS pool version
15. I''m
> playing around a bit to make it break.
> 
> Now I write some garbage to one of the log mirror devices.
> 
> # dd if=/dev/urandom of=/dev/zvol/dsk/data/LogMirrorDev2 bs=1024
> count=100000
Log devices are generally write-only.  They are only read during boot, after an
ungraceful crash.  It is extremely difficult to get a significant number of GB
used on the log device, because they are flushed out to primary storage so
frequently.  They are not read when you do a scrub.  For all these reasons, it
will be extremely difficult for you to experimentally produce a checksum error
on a log device.

On Nov 13, 2010, at 7:33 AM, Edward Ned Harvey wrote:
> Log devices are generally write-only.  They are only read during boot,
after an ungraceful crash.  It is extremely difficult to get a significant
number of GB used on the log device, because they are flushed out to primary
storage so frequently.  They are not read when you do a scrub.
This is wrong. We do read ZIL blocks when scrubbing, as you for example can have
a filesystem with non-empty ZIL which is not mounted.

regards
victor

See this thread - it dealt with the same question somehow.

  http://www.mail-archive.com/zfs-discuss at opensolaris.org/msg41715.html

Regards, 
Robert
-- 
This message posted from opensolaris.org