zfs discuss - May 2007 - ZFS Storage Pools Recommendations for Productive Environments

Hello,

solaris Internals wiki contains many interesting things about zfs.
But i have no glue about the reasons for this entry:

In Section "ZFS Storage Pools Recommendations - Storage Pools" you can
read:
[i]For all production environments, set up a redundant ZFS storage pool, such as
a raidz, raidz2, or a mirrored configuration, regardless of the RAID level
implemented on the underlying storage device.[/i]
(see
<http://www.solarisinternals.com/wiki/index.php/ZFS_Best_Practices_Guide#Storage_Pools>)

In our environment we use EMC based storage subsystems which are protected by
RAID1 (mirrored disks).
What''s the reason for building an upper level zfs mirror on these
already mirrored disks?

Or do I misinterpret solaris internals wiki?
 
 
This message posted from opensolaris.org

On 9-May-07, at 4:45 AM, Andreas Koppenhoefer wrote:
> Hello,
>
> solaris Internals wiki contains many interesting things about zfs.
> But i have no glue about the reasons for this entry:
>
> In Section "ZFS Storage Pools Recommendations - Storage Pools"
you
> can read:
> [i]For all production environments, set up a redundant ZFS storage  
> pool, such as a raidz, raidz2, or a mirrored configuration,  
> regardless of the RAID level implemented on the underlying storage  
> device.[/i]
> (see <http://www.solarisinternals.com/wiki/index.php/ 
> ZFS_Best_Practices_Guide#Storage_Pools>)
>
> In our environment we use EMC based storage subsystems which are  
> protected by RAID1 (mirrored disks).
> What''s the reason for building an upper level zfs mirror on these
> already mirrored disks?
It''s necessary if you wish ZFS to self-heal your data from errors  
detected in the underlying subsystems. Without redundancy at the pool  
level it can detect them (checksums) but not repair them.

--Toby
>
> Or do I misinterpret solaris internals wiki?
>
>
> This message posted from opensolaris.org
> _______________________________________________
> zfs-discuss mailing list
> zfs-discuss at opensolaris.org
> http://mail.opensolaris.org/mailman/listinfo/zfs-discuss

comment below...

Toby Thain wrote:
> 
> On 9-May-07, at 4:45 AM, Andreas Koppenhoefer wrote:
> 
>> Hello,
>>
>> solaris Internals wiki contains many interesting things about zfs.
>> But i have no glue about the reasons for this entry:
>>
>> In Section "ZFS Storage Pools Recommendations - Storage
Pools" you can
>> read:
>> [i]For all production environments, set up a redundant ZFS storage 
>> pool, such as a raidz, raidz2, or a mirrored configuration, regardless 
>> of the RAID level implemented on the underlying storage device.[/i]
>> (see 
>>
<http://www.solarisinternals.com/wiki/index.php/ZFS_Best_Practices_Guide#Storage_Pools>)
>>
>>
>> In our environment we use EMC based storage subsystems which are 
>> protected by RAID1 (mirrored disks).
>> What''s the reason for building an upper level zfs mirror on
these
>> already mirrored disks?
> 
> It''s necessary if you wish ZFS to self-heal your data from errors 
> detected in the underlying subsystems. Without redundancy at the pool 
> level it can detect them (checksums) but not repair them.
Yes, however with copies, it is not necessarily required that we build
a redundant zpool.  I''ll update the wiki.
  -- richard

which one is the most performant: copies=2 or zfs-mirror?

s.

On 5/9/07, Richard Elling <Richard.Elling at sun.com>
wrote:
> comment below...
>
> Toby Thain wrote:
> >
> > On 9-May-07, at 4:45 AM, Andreas Koppenhoefer wrote:
> >
> >> Hello,
> >>
> >> solaris Internals wiki contains many interesting things about zfs.
> >> But i have no glue about the reasons for this entry:
> >>
> >> In Section "ZFS Storage Pools Recommendations - Storage
Pools" you can
> >> read:
> >> [i]For all production environments, set up a redundant ZFS storage
> >> pool, such as a raidz, raidz2, or a mirrored configuration,
regardless
> >> of the RAID level implemented on the underlying storage
device.[/i]
> >> (see
> >>
<http://www.solarisinternals.com/wiki/index.php/ZFS_Best_Practices_Guide#Storage_Pools>)
> >>
> >>
> >> In our environment we use EMC based storage subsystems which are
> >> protected by RAID1 (mirrored disks).
> >> What''s the reason for building an upper level zfs mirror
on these
> >> already mirrored disks?
> >
> > It''s necessary if you wish ZFS to self-heal your data from
errors
> > detected in the underlying subsystems. Without redundancy at the pool
> > level it can detect them (checksums) but not repair them.
>
> Yes, however with copies, it is not necessarily required that we build
> a redundant zpool.  I''ll update the wiki.
>   -- richard
> _______________________________________________
> zfs-discuss mailing list
> zfs-discuss at opensolaris.org
> http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
>

> which one is the most performant: copies=2 or zfs-mirror?
What type of copies are you talking about?
Mirrored data in underlying storage subsystem or a (new) feature in zfs?

- Andreas
 
 
This message posted from opensolaris.org

Andreas Koppenhoefer wrote:
>> which one is the most performant: copies=2 or zfs-mirror?
Good question, hope to have some data soon.  From the back of the napkin
analysis, for the 2-disk case, it will be very similar.  However, copies
offers more possibilities than just 2 disks, so there is more interesting
work to be done.
> What type of copies are you talking about?
> Mirrored data in underlying storage subsystem or a (new) feature in zfs?
This is the "ditto blocks for data" feature of ZFS.  See
	http://blogs.sun.com/bill/entry/ditto_blocks_the_amazing_tape
	http://blogs.sun.com/relling/entry/zfs_copies_and_data_protection

  -- richard