zfs discuss - Jul 2007 - how to remove sun volume mgr configuration?

I''m brand new to zfs.

I have a system with a T3 array that I want to configure with zfs.
The T3 array had been setup with the Sun Volume manager in one big 250 
Gb volume.

I want to remove this volume and setup zfs.
My problem is the system has had a new OS installed (Sol10u3) since the 
volume was created.

The volume is mounted on my /export2 drive.

However smc does not see the volume.  It says I am not the owner.
The md.tab file for the volume information is empty. 

How do I get to the raw T3 disks so I can create the zfs system?
(I''m looking through both the Solaris Volume Manager Admin Guide and
the
ZFS Admin Guide. no luck.)

Russ
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://mail.opensolaris.org/pipermail/zfs-discuss/attachments/20070716/475442e2/attachment.html>

My meta* commands all return:
"... there are no existing databases"

This is the T3 array:

root at ps-eng8: format
Searching for disks...done
AVAILABLE DISK SELECTIONS:
 .....
       2. c1t1d0 <SUN-T300-0201 cyl 34530 alt 2 hd 128 sec 128>
          /ssm at 0,0/pci at 18,700000/SUNW,qlc at 2/fp at 0,0/ssd at
w50020f230000fb84,0

mount:
/export2 on /dev/dsk/c1t1d0s0 
read/write/setuid/devices/intr/largefiles/logging/xattr/onerror=panic/dev=1d80008
on Mon Jul 16 17:42:52 2007

I''m a little confused as to how the volume can be mounted and not be 
recognized by the meta commands.



Russ Petruzzelli wrote:
> I''m brand new to zfs.
>
> I have a system with a T3 array that I want to configure with zfs.
> The T3 array had been setup with the Sun Volume manager in one big 250 
> Gb volume.
>
> I want to remove this volume and setup zfs.
> My problem is the system has had a new OS installed (Sol10u3) since 
> the volume was created.
>
> The volume is mounted on my /export2 drive.
>
> However smc does not see the volume.  It says I am not the owner.
> The md.tab file for the volume information is empty. 
>
> How do I get to the raw T3 disks so I can create the zfs system?
> (I''m looking through both the Solaris Volume Manager Admin Guide
and
> the ZFS Admin Guide. no luck.)
>
> Russ
>
>------------------------------------------------------------------------
>
>_______________________________________________
>zfs-discuss mailing list
>zfs-discuss at opensolaris.org
>http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
>  
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://mail.opensolaris.org/pipermail/zfs-discuss/attachments/20070716/c744504a/attachment.html>

> My meta* commands all return:
> "... there are no existing databases"
Then you''re not using SVM volumes.
> This is the T3 array:
> 
> root at ps-eng8: format
> Searching for disks...done
> AVAILABLE DISK SELECTIONS:
>  .....
>        2. c1t1d0 <SUN-T300-0201 cyl 34530 alt 2 hd 128 sec 128>
>           /ssm at 0,0/pci at 18,700000/SUNW,qlc at 2/fp at 0,0/ssd at
w50020f230000fb84,0
> 
> mount:
> /export2 on /dev/dsk/c1t1d0s0 
>
read/write/setuid/devices/intr/largefiles/logging/xattr/onerror=panic/dev=1d80008
> on Mon Jul 16 17:42:52 2007
That''s a raw T3 volume, with UFS mounted directly.
> I''m a little confused as to how the volume can be mounted and not
be
> recognized by the meta commands.
The T3 has an internal raid controller.  I''m pretty sure the first
versions had no way to access the raw disks.  You could create up to 2
luns, either R5 or some type of mirrored setup.  I don''t know which
versions allowed more than that.

It''s not a JBOD, and you may not be able to use it as one.

You can use ZFS on that volume, but it will have no redundancy at the
ZFS level, only at the disk level controlled by the T3.

-- 
Darren Dunham                                           ddunham at taos.com
Senior Technical Consultant         TAOS            http://www.taos.com/
Got some Dr Pepper?                           San Francisco, CA bay area
         < This line left intentionally blank to confuse you. >

Thanks Darren,

    re:  "You can use ZFS on that volume, but it will have no redundancy
    at the ZFS level, only at the disk level controlled by the T3."

I believe it is an older T3.  

I''m using this system in a test lab, so data integrity is not too
important for me.  I mainly want to see what kind of performance I can get out
of the zfs/T3 setup.

I do see a note on pg 31 of the zfs admin guide that recommends against this
configuration. (but saying it is possible).

So would I get any performance bump using zfs on the T3 volume?

Or am I just getting myself into shark infested waters?

Russ


Darren Dunham wrote:
>>My meta* commands all return:
>>"... there are no existing databases"
>>    
>>
>
>Then you''re not using SVM volumes.
>
>  
>
>>This is the T3 array:
>>
>>root at ps-eng8: format
>>Searching for disks...done
>>AVAILABLE DISK SELECTIONS:
>> .....
>>       2. c1t1d0 <SUN-T300-0201 cyl 34530 alt 2 hd 128 sec 128>
>>          /ssm at 0,0/pci at 18,700000/SUNW,qlc at 2/fp at 0,0/ssd at
w50020f230000fb84,0
>>
>>mount:
>>/export2 on /dev/dsk/c1t1d0s0 
>>read/write/setuid/devices/intr/largefiles/logging/xattr/onerror=panic/dev=1d80008
>>on Mon Jul 16 17:42:52 2007
>>    
>>
>
>That''s a raw T3 volume, with UFS mounted directly.
>
>  
>
>>I''m a little confused as to how the volume can be mounted and
not be
>>recognized by the meta commands.
>>    
>>
>
>The T3 has an internal raid controller.  I''m pretty sure the first
>versions had no way to access the raw disks.  You could create up to 2
>luns, either R5 or some type of mirrored setup.  I don''t know which
>versions allowed more than that.
>
>It''s not a JBOD, and you may not be able to use it as one.
>
>You can use ZFS on that volume, but it will have no redundancy at the
>ZFS level, only at the disk level controlled by the T3.
>
>  
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://mail.opensolaris.org/pipermail/zfs-discuss/attachments/20070716/ac8c74e7/attachment.html>

> I''m using this system in a test lab, so data integrity is not too
> important for me.  I mainly want to see what kind of performance I can
> get out of the zfs/T3 setup.
> I do see a note on pg 31 of the zfs admin guide that recommends against
this configuration. (but saying it is possible).
> 
> So would I get any performance bump using zfs on the T3 volume?
Compared to running UFS on it directly?  I wouln''t think
there''d be a
significant difference, but for some workloads you might.  ZFS will
schedule writes differently than UFS will.  That might improve some
things.
> Or am I just getting myself into shark infested waters?
Not to my mind.  It''s just that the configuration you have
doesn''t allow
some of the strengths of ZFS to be used well.

-- 
Darren Dunham                                           ddunham at taos.com
Senior Technical Consultant         TAOS            http://www.taos.com/
Got some Dr Pepper?                           San Francisco, CA bay area
         < This line left intentionally blank to confuse you. >

On Mon, 2007-07-16 at 18:19 -0700, Russ Petruzzelli
wrote:
> Or am I just getting myself into shark infested waters?
configurations that might be interesting to play with:
(emphasis here on "play"...)

 1) use the T3''s management CLI to reconfigure the T3 into two raid-0
volumes, and mirror them with ZFS.  

 2) if you have some JBODs available as well, use the T3 (which has a
modest-sized battery backed write cache in the controller) as a separate
log device (that''s a new feature introduced in a recent nevada build).

					- Bill

Russ Petruzzelli wrote:
>    Thanks Darren,
> 
>     re:  " You can use ZFS on that volume, but it will have no
>     redundancy at the ZFS level, only at the disk level controlled by
>     the T3."
> 
> I believe it is an older T3.  
Performance-wise, these are pretty wimpy.  You should be able to saturate
the array controller, even without enabling RAID-5 on it.  Note that the
T3''s implementation of RAID-0 isn''t quite the same as other
arrays, so it
may perform somewhat differently (better) than you expect from JBODs.
  -- richard

Bill Sommerfeld wrote:
> On Mon, 2007-07-16 at 18:19 -0700, Russ Petruzzelli wrote:
>   
>> Or am I just getting myself into shark infested waters?
>>     
>
> configurations that might be interesting to play with:
> (emphasis here on "play"...)
>
>  1) use the T3''s management CLI to reconfigure the T3 into two
raid-0
> volumes, and mirror them with ZFS.  
>
>  2) if you have some JBODs available as well, use the T3 (which has a
> modest-sized battery backed write cache in the controller) as a separate
> log device (that''s a new feature introduced in a recent nevada
build).
Has the project that lets you specify an array has having a battery back 
up gone in yet? If not then wouldn''t the sync cache problem be in play?
I don''t know if the T3 honors cache flush commands or set the
"i''ve got
a stable cache" bit in the relevant scsi mode page.

Darren Dunham wrote:
>> My meta* commands all return:
>> "... there are no existing databases"
> Then you''re not using SVM volumes.
Correct. No metadb, no SVM.
>> This is the T3 array:
>>
>> root at ps-eng8: format
>> Searching for disks...done
>> AVAILABLE DISK SELECTIONS:
>>  .....
>>        2. c1t1d0 <SUN-T300-0201 cyl 34530 alt 2 hd 128 sec 128>
>>           /ssm at 0,0/pci at 18,700000/SUNW,qlc at 2/fp at 0,0/ssd at
w50020f230000fb84,0
>>
>> mount:
>> /export2 on /dev/dsk/c1t1d0s0 
>>
read/write/setuid/devices/intr/largefiles/logging/xattr/onerror=panic/dev=1d80008
>> on Mon Jul 16 17:42:52 2007
> 
> That''s a raw T3 volume, with UFS mounted directly.
> 
>> I''m a little confused as to how the volume can be mounted and
not be
>> recognized by the meta commands.
> 
> The T3 has an internal raid controller.  I''m pretty sure the first
> versions had no way to access the raw disks.  You could create up to 2
> luns, either R5 or some type of mirrored setup.  I don''t know
which
> versions allowed more than that.
The T3B with fw v3.x (I think) and the T4 (aka 6020 tray) allow
more than two volumes, but you''re still quite restricted in what
you can do with them.
> It''s not a JBOD, and you may not be able to use it as one.
You are definitely not able to use a T3 as a jbod.
> You can use ZFS on that volume, but it will have no redundancy at the
> ZFS level, only at the disk level controlled by the T3.
Well ... you could create two volumes on the array and mirror
those using ZFS .... Some might say that''s a waste of space :)


James C. McPherson
--
Solaris kernel software engineer, system admin and troubleshooter
               http://www.jmcp.homeunix.com/blog
Find me on LinkedIn @ http://www.linkedin.com/in/jamescmcpherson

James C. McPherson wrote:
>
>
> The T3B with fw v3.x (I think) and the T4 (aka 6020 tray) allow
> more than two volumes, but you''re still quite restricted in what
> you can do with them.
>   
You are limited to two raid groups with slices on top of those raid 
groups presented as LUNs. I''d just stick with the raid groups and not 
try to get really overboard and use slices because you can just ....
>   
> [SNIP]
>
>   
>> You can use ZFS on that volume, but it will have no redundancy at the
>> ZFS level, only at the disk level controlled by the T3.
>>     
>
> Well ... you could create two volumes on the array and mirror
> those using ZFS .... Some might say that''s a waste of space :)

... stick to R0 and then mirror with ZFS? At least T3 will let you do 
that as opposed to other storage arrays that let you pick from R1 and R5 
only.

On 7/17/07, Richard Elling <Richard.Elling at sun.com>
wrote:
> Performance-wise, these are pretty wimpy.  You should be able to saturate
> the array controller, even without enabling RAID-5 on it.  Note that the
> T3''s implementation of RAID-0 isn''t quite the same as
other arrays, so it
> may perform somewhat differently (better) than you expect from JBODs.
This is similiar something I intend to do when we finally move off our
T3''s -- we have 9 T3, 2 T3B.

I''d create 2 LUNs, 5 Disk RAID0 + 4 Disk RAID0 on each of the T3 and
stagger mirror the LUNs with ZFS.  The final pool will be used to
house user filesystems for a compute cluster.  I haven''t gotten any
benchmarks on this config yet -- still waiting for my T3''s to come
along...

Can you please share the difference with the T3 implementation of RAID0?



-- 
Just me,
Wire ...
Blog: <prstat.blogspot.com>

Hi Wee Yeh,

Thanks for the earlier tips on July17th.  I have a couple questions...

I followed your suggestion and first created two raid0 volumes.

    volume         capacity     raid  data      standby
    v0             134.890 GB   0     u1d1-4    none
    v1             168.613 GB   0     u1d5-9    none


I wasn''t sure whether I should:
1)  create zones in the ufs / file system and then create zfs file 
systems in the zones,
or...
2)  create the zfs file system and its heirarchy and then put zones in 
the zfs systems.

I did #2.


And then created the zfs mirrors:
 >#zpool create T3arraymirror mirror c1t1d0 c1t1d1

root at ps-eng8: zpool list
NAME                    SIZE    USED   AVAIL    CAP  HEALTH     ALTROOT
T3arraymirror           134G     80K    134G     0%  ONLINE     -

root at ps-eng8: zfs create zpool1/zone1
root at ps-eng8: zfs create T3arraymirror/zones
root at ps-eng8: zfs create T3arraymirror/zones/zone2
root at ps-eng8:  zfs create T3arraymirror/zones/zone3
root at ps-eng8:  zfs create T3arraymirror/zones/zone4
root at ps-eng8:  zfs create T3arraymirror/zones/zone5

And then created a zone with zonemgr.

Will my zone I created inherit a zfs file system since it was created in 
one?

Does it make sense?  It''s my first try at this....

Thanks,
Russ



Wee Yeh Tan wrote:
> On 7/17/07, Richard Elling <Richard.Elling at sun.com> wrote:
>
>> Performance-wise, these are pretty wimpy.  You should be able to 
>> saturate
>> the array controller, even without enabling RAID-5 on it.  Note that
the
>> T3''s implementation of RAID-0 isn''t quite the same as
other arrays,
>> so it
>> may perform somewhat differently (better) than you expect from JBODs.
>
>
> This is similiar something I intend to do when we finally move off our
> T3''s -- we have 9 T3, 2 T3B.
>
> I''d create 2 LUNs, 5 Disk RAID0 + 4 Disk RAID0 on each of the T3
and
> stagger mirror the LUNs with ZFS.  The final pool will be used to
> house user filesystems for a compute cluster.  I haven''t gotten
any
> benchmarks on this config yet -- still waiting for my T3''s to come
> along...
>
> Can you please share the difference with the T3 implementation of RAID0?
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://mail.opensolaris.org/pipermail/zfs-discuss/attachments/20070726/8ac0d10b/attachment.html>