zfs discuss - Oct 2006 - ZFS Usability issue : improve means of finding ZFS<->physdevice(s) mapping

ZFS is supposed to be much easier to use than UFS.

For creating a filesystem, I agree it is, as I could do that easily without a
man page.

However, I found it rather surprising that I could not see the physical
device(s) a zfs filesystem was attached to using either "df" command
(that shows physical device mount points for all other file systems), or even
the "zfs" command.

Even going to "zpool" command it took a few minutes to finally stumble
across the only two commands that will give you that information, as it is not
exactly intuitive.

Ideally, I''d think "df" should show physical device
connections of zfs pools, though I can imagine there may be some circumstances
where that is not desirable so perhaps a new argument would be needed to specify
if that detail is shown or not.

If this is not done, I think "zfs list -v"  (-v is not currently an
option to the zfs list command) should show the physical devices in use by the
pools.

In any case, I think it is clear "zpool list" should have a
"-v" argument added that will show the device associations, so that
people don''t have to stumble blindly until they run into the
"zpool iostat -v" or "zpool status -v" commands to finally
accomplish this rather simple task.

Any comments on the above?  I''m using S10 06/06, so perhaps
I''ll get lucky and someone has already added one or all the above
improvements. :)

Cheers,

   Bruce
 
 
This message posted from opensolaris.org

I  don''t understand why you can''t use ''zpool
status''?  That will show
the pools and the physical devices in each and is also a pretty basic  
command.  Examples are given in the sysadmin docs and manpages for  
ZFS on the opensolaris ZFS community page.

I realize it''s not quite the same command as in UFS, and it''s
easier
when things remain the same, but it''s a different filesystem so you  
need some different commands that make more sense for how it''s  
structured. The idea being hopefully that  soon zpool and zfs  
commands will become just as ''intuitive'' for people :)

Noel

(p.s. not to mention am I the only person that thinks that ''zpool  
status'' (in human speak, not geek) makes more sense than
''df''? wtf )

On Oct 13, 2006, at 1:55 PM, Bruce Chapman wrote:
> ZFS is supposed to be much easier to use than UFS.
>
> For creating a filesystem, I agree it is, as I could do that easily  
> without a man page.
>
> However, I found it rather surprising that I could not see the  
> physical device(s) a zfs filesystem was attached to using either  
> "df" command (that shows physical device mount points for all
other
> file systems), or even the "zfs" command.
>
> Even going to "zpool" command it took a few minutes to finally  
> stumble across the only two commands that will give you that  
> information, as it is not exactly intuitive.
>
> Ideally, I''d think "df" should show physical device
connections of
> zfs pools, though I can imagine there may be some circumstances  
> where that is not desirable so perhaps a new argument would be  
> needed to specify if that detail is shown or not.
>
> If this is not done, I think "zfs list -v"  (-v is not currently
an
> option to the zfs list command) should show the physical devices in  
> use by the pools.
>
> In any case, I think it is clear "zpool list" should have a
"-v"
> argument added that will show the device associations, so that  
> people don''t have to stumble blindly until they run into the
"zpool
> iostat -v" or "zpool status -v" commands to finally
accomplish this
> rather simple task.
>
> Any comments on the above?  I''m using S10 06/06, so perhaps
I''ll
> get lucky and someone has already added one or all the above  
> improvements. :)
>
> Cheers,
>
>    Bruce
>
>
> This message posted from opensolaris.org
> _______________________________________________
> zfs-discuss mailing list
> zfs-discuss at opensolaris.org
> http://mail.opensolaris.org/mailman/listinfo/zfs-discuss

Hello Noel,

Friday, October 13, 2006, 11:22:06 PM, you wrote:

ND> I  don''t understand why you can''t use ''zpool
status''?  That will show
ND> the pools and the physical devices in each and is also a pretty basic
ND> command.  Examples are given in the sysadmin docs and manpages for  
ND> ZFS on the opensolaris ZFS community page.

Showing physical devs in df output with ZFS is not right and I do not
imagine how one would show in df output for a pool with dozen disks.

But an option to zpool command to display config in such a way so it''s
easy (almost copy&paste) to recreate such config would be useful.
Something like metastat -p.



-- 
Best regards,
 Robert                            mailto:rmilkowski at task.gda.pl
                                       http://milek.blogspot.com

Robert Milkowski wrote:
> Hello Noel,
> 
> Friday, October 13, 2006, 11:22:06 PM, you wrote:
> 
> ND> I  don''t understand why you can''t use
''zpool status''?  That will show
> ND> the pools and the physical devices in each and is also a pretty
basic
> ND> command.  Examples are given in the sysadmin docs and manpages for  
> ND> ZFS on the opensolaris ZFS community page.
> 
> Showing physical devs in df output with ZFS is not right and I do not
> imagine how one would show in df output for a pool with dozen disks.
> 
> But an option to zpool command to display config in such a way so
it''s
> easy (almost copy&paste) to recreate such config would be useful.
> Something like metastat -p.
Agreed, see 6276640 "zpool config".

--matt

On Oct 13, 2006, at 4:55 PM, Bruce Chapman wrote:
> ZFS is supposed to be much easier to use than UFS.
>
> For creating a filesystem, I agree it is, as I could do that easily  
> without a man page.
>
> However, I found it rather surprising that I could not see the  
> physical device(s) a zfs filesystem was attached to using either  
> "df" command (that shows physical device mount points for all
other
> file systems), or even the "zfs" command.
>
> Even going to "zpool" command it took a few minutes to finally  
> stumble across the only two commands that will give you that  
> information, as it is not exactly intuitive.
>
> Ideally, I''d think "df" should show physical device
connections of
> zfs pools, though I can imagine there may be some circumstances  
> where that is not desirable so perhaps a new argument would be  
> needed to specify if that detail is shown or not.
How is this different from, say, SVM and its representation of  
metadevices? You don''t see the underlying slices which make up
md''s
in df output... those are abstracted away. ZFS in not different from  
this on a conceptual level here.

Just like you would use ''metastat -c'' to see the underlying
physical
devices there, you would use ''zpool status'' in ZFS. Your
example ''zfs
list -v'' command would essentially be redundant in the face of the  
existing ''zpool status'' anyway...

/dale

Noel Dellofano wrote:
> I  don''t understand why you can''t use ''zpool
status''?  That will show
> the pools and the physical devices in each and is also a pretty basic 
> command.  Examples are given in the sysadmin docs and manpages for ZFS 
> on the opensolaris ZFS community page.
> 
> I realize it''s not quite the same command as in UFS, and
it''s easier
> when things remain the same, but it''s a different filesystem so
you need
> some different commands that make more sense for how it''s
structured.
> The idea being hopefully that  soon zpool and zfs commands will become 
> just as ''intuitive'' for people :)
I agree.  Whats more with UFS it is very often UFS+SVM so what you see 
in df is SVM metadevices not real disks either.  In the UFS+SVM case you 
need to use metastat(1M) and work your way through that output to find 
the actual physical disk.

IMO the whole point of ZFS was abstracting this physical disk thing away 
from you when you don''t need to use it, and df output IMO is one place 
it really doesn''t belong.

-- 
Darren J Moffat

On Oct 16, 2006, at 07:39, Darren J Moffat wrote:
> Noel Dellofano wrote:
>> I  don''t understand why you can''t use ''zpool
status''?  That will
>> show the pools and the physical devices in each and is also a  
>> pretty basic command.  Examples are given in the sysadmin docs and  
>> manpages for ZFS on the opensolaris ZFS community page.
>> I realize it''s not quite the same command as in UFS, and
it''s
>> easier when things remain the same, but it''s a different  
>> filesystem so you need some different commands that make more  
>> sense for how it''s structured. The idea being hopefully that 
soon
>> zpool and zfs commands will become just as
''intuitive'' for people :)
>
> I agree.  Whats more with UFS it is very often UFS+SVM so what you  
> see in df is SVM metadevices not real disks either.  In the UFS+SVM  
> case you need to use metastat(1M) and work your way through that  
> output to find the actual physical disk.
>
> IMO the whole point of ZFS was abstracting this physical disk thing  
> away from you when you don''t need to use it, and df output IMO is
> one place it really doesn''t belong.
I agree, and on a deeper level it''s simply a matter of which level of  
virtualization you want to see and where you want to see it.  We  
could argue that the devfs tree shows more of the physical device  
than the links in /dev, but the syntax of both of these is somewhat  
obscure and we''ve just learned to deal with the c#t#d# model/ 
structure.  When we get into large scale environments and unique  
target identifiers - MPxIO or STMS introduced the idea of putting a  
GUID in the target string, but this becomes somewhat unwieldy for  
many situations since it can often be unclear what storage [device,  
port, vdisk, lun] "60060E80047D410000007D4100000408" is really mapped
to.  For this you typically have to resort to luxadm(1M) and cfgadm_fp 
(1M) commands to determine which device went where.

Then if you''re doing storage based volume management, you''d
need some
tool to communicate with the array to figure out which virtual disk  
is composed of which underlying physical devices.  And if you use a  
host-based volume management you would need to use those tools as  
well (metastat, vxdisk, etc) to see more into the physical structure  
of the volumes that were created.  Hopefully you''ve chosen  
intelligent names for things, and I would argue that interacting with  
an arbitrary name-based structure, or the abstracted structure, has  
more of an advantage in reducing some of the underlying visual  
complexity.

One would hope that in a few years we would move more toward an SNS  
or OSD type approach with devices accessible through a name lookup or  
a logical OST name and we could finally move beyond the SCSI  
numbering approach with everything displayed in the raw, but I  
believe that we''re still a long way off.

Now df simply maps from your mount table - so I don''t see either the  
df options being extended nor the zfs/zpool interface changing ..  
rather I see this as an opportunity for us to use more logical and  
diverse naming structures for groupings of underlying devices with  
new tools to manipulate both.

.je