Hi, I am soliciting input from the ZFS engineers and/or ZFS users on an extension to "zfs list". Thanks in advance for your feedback. Quick Background: The pNFS project (http://opensolaris.org/os/project/nfsv41/) is adding a new DMU object set type which is used on the pNFS data server to store pNFS stripe DMU objects. A pNFS dataset gets created with the "zfs create" command and gets displayed using "zfs list". Specific Question: Should the pNFS datasets show up in the default "zfs list" output? Just as with ZFS file systems and ZVOLs, the number of pNFS datasets that exist on a data server will vary depending on the configuration. The following is output from the modified command and reflects the current mode of operation (i.e. "zfs list" lists filesystems, volumes and pnfs datasets by default): (pnfs-17-21:/home/lisagab):6 % zfs list NAME USED AVAIL REFER MOUNTPOINT rpool 30.0G 37.0G 32.5K /rpool rpool/ROOT 18.2G 37.0G 18K legacy rpool/ROOT/snv_105 18.2G 37.0G 6.86G / rpool/ROOT/snv_105/var 11.4G 37.0G 11.4G /var rpool/dump 9.77G 37.0G 9.77G - rpool/export 40K 37.0G 21K /export rpool/export/home 19K 37.0G 19K /export/home rpool/pnfsds 31K 37.0G 15K - <---pNFS dataset rpool/pnfsds/47C80414080A4A42 16K 37.0G 16K - <---pNFS dataset rpool/swap 1.97G 38.9G 4.40M - (pnfs-17-21:/home/lisagab):7 % zfs list -t pnfsdata NAME USED AVAIL REFER MOUNTPOINT rpool/pnfsds 31K 37.0G 15K - rpool/pnfsds/47C80414080A4A42 16K 37.0G 16K - (pnfs-17-21:/home/lisagab):8 % zfs get all rpool/pnfsds NAME PROPERTY VALUE SOURCE rpool/pnfsds type pnfs-dataset - rpool/pnfsds creation Mon Feb 2 13:56 2009 - rpool/pnfsds used 31K - rpool/pnfsds available 37.0G - rpool/pnfsds referenced 15K - rpool/pnfsds compressratio 1.00x - rpool/pnfsds quota none default rpool/pnfsds reservation none default rpool/pnfsds recordsize 128K default rpool/pnfsds checksum on default rpool/pnfsds compression off default rpool/pnfsds zoned off default rpool/pnfsds copies 1 default rpool/pnfsds refquota none default rpool/pnfsds refreservation none default rpool/pnfsds sharepnfs off default rpool/pnfsds mds none default Thanks, Lisa -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://mail.opensolaris.org/pipermail/zfs-discuss/attachments/20090302/9812a9d5/attachment.html>
Yes. -- richard Lisa Week wrote:> Hi, > > I am soliciting input from the ZFS engineers and/or ZFS users on an > extension to "zfs list". Thanks in advance for your feedback. > > Quick Background: > The pNFS project (http://opensolaris.org/os/project/nfsv41/) > <http://opensolaris.org/os/project/nfsv41/%29> is adding a new DMU > object set type which is used on the pNFS data server to store pNFS > stripe DMU objects. A pNFS dataset gets created with the "zfs create" > command and gets displayed using "zfs list". > > Specific Question: > Should the pNFS datasets show up in the default "zfs list" output? > Just as with ZFS file systems and ZVOLs, the number of pNFS datasets > that exist on a data server will vary depending on the configuration. > > The following is output from the modified command and reflects the > current mode of operation (i.e. "zfs list" lists filesystems, volumes > and pnfs datasets by default): > > (pnfs-17-21:/home/lisagab):6 % zfs list > NAME USED AVAIL REFER MOUNTPOINT > rpool 30.0G 37.0G 32.5K /rpool > rpool/ROOT 18.2G 37.0G 18K legacy > rpool/ROOT/snv_105 18.2G 37.0G 6.86G / > rpool/ROOT/snv_105/var 11.4G 37.0G 11.4G /var > rpool/dump 9.77G 37.0G 9.77G - > rpool/export 40K 37.0G 21K /export > rpool/export/home 19K 37.0G 19K /export/home > rpool/pnfsds 31K 37.0G 15K - <---pNFS > dataset > rpool/pnfsds/47C80414080A4A42 16K 37.0G 16K - <---pNFS > dataset > rpool/swap 1.97G 38.9G 4.40M - > > (pnfs-17-21:/home/lisagab):7 % zfs list -t pnfsdata > NAME USED AVAIL REFER MOUNTPOINT > rpool/pnfsds 31K 37.0G 15K - > rpool/pnfsds/47C80414080A4A42 16K 37.0G 16K - > > (pnfs-17-21:/home/lisagab):8 % zfs get all rpool/pnfsds > NAME PROPERTY VALUE SOURCE > rpool/pnfsds type pnfs-dataset - > rpool/pnfsds creation Mon Feb 2 13:56 2009 - > rpool/pnfsds used 31K - > rpool/pnfsds available 37.0G - > rpool/pnfsds referenced 15K - > rpool/pnfsds compressratio 1.00x - > rpool/pnfsds quota none default > rpool/pnfsds reservation none default > rpool/pnfsds recordsize 128K default > rpool/pnfsds checksum on default > rpool/pnfsds compression off default > rpool/pnfsds zoned off default > rpool/pnfsds copies 1 default > rpool/pnfsds refquota none default > rpool/pnfsds refreservation none default > rpool/pnfsds sharepnfs off default > rpool/pnfsds mds none default > > Thanks, > Lisa > > ------------------------------------------------------------------------ > > _______________________________________________ > zfs-discuss mailing list > zfs-discuss at opensolaris.org > http://mail.opensolaris.org/mailman/listinfo/zfs-discuss >
Lisa, On Mon, Mar 02, 2009 at 09:58:08PM -0700, Lisa Week wrote:> Should the pNFS datasets show up in the default "zfs list" output? Just > as with ZFS file systems and ZVOLs, the number of pNFS datasets that > exist on a data server will vary depending on the configuration.I think they should be listed. Thanks and regards, Sanjeev> > The following is output from the modified command and reflects the > current mode of operation (i.e. "zfs list" lists filesystems, volumes > and pnfs datasets by default): > > (pnfs-17-21:/home/lisagab):6 % zfs list > NAME USED AVAIL REFER MOUNTPOINT > rpool 30.0G 37.0G 32.5K /rpool > rpool/ROOT 18.2G 37.0G 18K legacy > rpool/ROOT/snv_105 18.2G 37.0G 6.86G / > rpool/ROOT/snv_105/var 11.4G 37.0G 11.4G /var > rpool/dump 9.77G 37.0G 9.77G - > rpool/export 40K 37.0G 21K /export > rpool/export/home 19K 37.0G 19K /export/home > rpool/pnfsds 31K 37.0G 15K - <---pNFS > dataset > rpool/pnfsds/47C80414080A4A42 16K 37.0G 16K - <---pNFS > dataset > rpool/swap 1.97G 38.9G 4.40M - > > (pnfs-17-21:/home/lisagab):7 % zfs list -t pnfsdata > NAME USED AVAIL REFER MOUNTPOINT > rpool/pnfsds 31K 37.0G 15K - > rpool/pnfsds/47C80414080A4A42 16K 37.0G 16K - > > (pnfs-17-21:/home/lisagab):8 % zfs get all rpool/pnfsds > NAME PROPERTY VALUE SOURCE > rpool/pnfsds type pnfs-dataset - > rpool/pnfsds creation Mon Feb 2 13:56 2009 - > rpool/pnfsds used 31K - > rpool/pnfsds available 37.0G - > rpool/pnfsds referenced 15K - > rpool/pnfsds compressratio 1.00x - > rpool/pnfsds quota none default > rpool/pnfsds reservation none default > rpool/pnfsds recordsize 128K default > rpool/pnfsds checksum on default > rpool/pnfsds compression off default > rpool/pnfsds zoned off default > rpool/pnfsds copies 1 default > rpool/pnfsds refquota none default > rpool/pnfsds refreservation none default > rpool/pnfsds sharepnfs off default > rpool/pnfsds mds none default > > Thanks, > Lisa >-- ---------------- Sanjeev Bagewadi Solaris RPE Bangalore, India
Richard Morris - Sun Microsystems - Burlington United States
2009-Mar-04 16:30 UTC
[zfs-discuss] zfs list extentions related to pNFS
> Quick Background: > The pNFS project (http://opensolaris.org/os/project/nfsv41/) > <http://opensolaris.org/os/project/nfsv41/%29> is adding a new DMU > object set type which is used on the pNFS data server to store pNFS > stripe DMU objects. A pNFS dataset gets created with the "zfs create" > command and gets displayed using "zfs list". > > Specific Question: > Should the pNFS datasets show up in the default "zfs list" output? > Just as with ZFS file systems and ZVOLs, the number of pNFS datasets > that exist on a data server will vary depending on the configuration.Lisa, In Solaris 10, ZFS datasets of all types show up in the default output of "zfs list". In Nevada, "zfs list" only shows by default datasets that are filesystems or volumes. Snapshots of filesystems or snapshots of volumes are not listed unless the pool level property called "listsnapshots", which is "off" by default, is set to "on". So the answer to the specific question above is, in Nevada, pNFS datasets should show up in the default output from "zfs list" but that snapshots of pNFS datasets should not unless "listsnapshots" is on. However, with the addition of this new dataset type, it will be difficult to distinguish a ZFS volume from a pNFS dataset. So perhaps the default output of "zfs list" should also include the dataset type. Maybe something like this: NAME TYPE USED AVAIL REFER MOUNTPOINT pool/fs FS 9.77G 37.0G 9.77G /pool/fs pool/vol ZVOL 512M 37.0G 16K - pool/pnfs pNFS 31K 37.0G 15K - -- Rich> The following is output from the modified command and reflects the > current mode of operation (i.e. "zfs list" lists filesystems, volumes > and pnfs datasets by default): > > (pnfs-17-21:/home/lisagab):6 % zfs list > NAME USED AVAIL REFER MOUNTPOINT > rpool 30.0G 37.0G 32.5K /rpool > rpool/ROOT 18.2G 37.0G 18K legacy > rpool/ROOT/snv_105 18.2G 37.0G 6.86G / > rpool/ROOT/snv_105/var 11.4G 37.0G 11.4G /var > rpool/dump 9.77G 37.0G 9.77G - > rpool/export 40K 37.0G 21K /export > rpool/export/home 19K 37.0G 19K /export/home > rpool/pnfsds 31K 37.0G 15K - <---pNFS > dataset > rpool/pnfsds/47C80414080A4A42 16K 37.0G 16K - <---pNFS > dataset > rpool/swap 1.97G 38.9G 4.40M - > > (pnfs-17-21:/home/lisagab):7 % zfs list -t pnfsdata > NAME USED AVAIL REFER MOUNTPOINT > rpool/pnfsds 31K 37.0G 15K - > rpool/pnfsds/47C80414080A4A42 16K 37.0G 16K - > > (pnfs-17-21:/home/lisagab):8 % zfs get all rpool/pnfsds > NAME PROPERTY VALUE SOURCE > rpool/pnfsds type pnfs-dataset - > rpool/pnfsds creation Mon Feb 2 13:56 2009 - > rpool/pnfsds used 31K - > rpool/pnfsds available 37.0G - > rpool/pnfsds referenced 15K - > rpool/pnfsds compressratio 1.00x - > rpool/pnfsds quota none default > rpool/pnfsds reservation none default > rpool/pnfsds recordsize 128K default > rpool/pnfsds checksum on default > rpool/pnfsds compression off default > rpool/pnfsds zoned off default > rpool/pnfsds copies 1 default > rpool/pnfsds refquota none default > rpool/pnfsds refreservation none default > rpool/pnfsds sharepnfs off default > rpool/pnfsds mds none default > > Thanks, > Lisa > > ------------------------------------------------------------------------ > > _______________________________________________ > 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/20090304/e0f5dde9/attachment.html>
Hi Rich, On Mar 4, 2009, at 9:30 AM, Richard Morris - Sun Microsystems - Burlington United States wrote:> >> Quick Background: >> The pNFS project (http://opensolaris.org/os/project/nfsv41/) is >> adding a new DMU object set type which is used on the pNFS data >> server to store pNFS stripe DMU objects. A pNFS dataset gets >> created with the "zfs create" command and gets displayed using "zfs >> list". >> >> Specific Question: >> Should the pNFS datasets show up in the default "zfs list" output? >> Just as with ZFS file systems and ZVOLs, the number of pNFS >> datasets that exist on a data server will vary depending on the >> configuration. > > Lisa, > > In Solaris 10, ZFS datasets of all types show up in the default > output of "zfs list". In Nevada, "zfs list" only shows by default > datasets that are filesystems or volumes. Snapshots of filesystems > or snapshots of volumes are not listed unless the pool level > property called "listsnapshots", which is "off" by default, is set > to "on". > > So the answer to the specific question above is, in Nevada, pNFS > datasets should show up in the default output from "zfs list" but > that snapshots of pNFS datasets should not unless "listsnapshots" is > on.Thanks and I agree about when to (and not to) display snapshots.> > > However, with the addition of this new dataset type, it will be > difficult to distinguish a ZFS volume from a pNFS dataset. So > perhaps the default output of "zfs list" should also include the > dataset type. Maybe something like this: > > NAME TYPE USED AVAIL REFER MOUNTPOINT > pool/fs FS 9.77G 37.0G 9.77G /pool/fs > pool/vol ZVOL 512M 37.0G 16K - > pool/pnfs pNFS 31K 37.0G 15K -That is good feedback. Thanks. It is not perfect, but you can do this, currently: (pnfs-17-21:/home/lisagab):6 % zfs list -o name,type,used,avail,refer,mountpoint NAME TYPE USED AVAIL REFER MOUNTPOINT rpool filesystem 30.0G 37.0G 32.5K /rpool rpool/ROOT filesystem 18.2G 37.0G 18K legacy rpool/ROOT/snv_105 filesystem 18.2G 37.0G 6.86G / rpool/ROOT/snv_105/var filesystem 11.4G 37.0G 11.4G /var rpool/dump volume 9.77G 37.0G 9.77G - rpool/export filesystem 40K 37.0G 21K / export rpool/export/home filesystem 19K 37.0G 19K / export/home rpool/pnfsds pnfs-dataset 31K 37.0G 15K - rpool/pnfsds/47C80414080A4A42 pnfs-dataset 16K 37.0G 16K - rpool/swap volume 1.97G 38.9G 4.40M - Having the above be the default "zfs list" output is a simple change to make, but I can''t commit do doing it unless, of course, there is buy in from the ZFS team. Does anyone have insight into any known problems it may cause to add the "type" property to the default output or why it was left out in the first place? Thanks, Lisa -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://mail.opensolaris.org/pipermail/zfs-discuss/attachments/20090304/cabe5cc7/attachment.html>
On Wed, Mar 04, 2009 at 02:13:51PM -0700, Lisa Week wrote:> (pnfs-17-21:/home/lisagab):6 % zfs list -o > name,type,used,avail,refer,mountpoint > NAME TYPE USED AVAIL REFER > MOUNTPOINT > rpool filesystem 30.0G 37.0G 32.5K /rpool > rpool/ROOT filesystem 18.2G 37.0G 18K legacy > rpool/ROOT/snv_105 filesystem 18.2G 37.0G 6.86G / > rpool/ROOT/snv_105/var filesystem 11.4G 37.0G 11.4G /var > rpool/dump volume 9.77G 37.0G 9.77G - > rpool/export filesystem 40K 37.0G 21K / > export > rpool/export/home filesystem 19K 37.0G 19K / > export/home > rpool/pnfsds pnfs-dataset 31K 37.0G 15K - > rpool/pnfsds/47C80414080A4A42 pnfs-dataset 16K 37.0G 16K - > rpool/swap volume 1.97G 38.9G 4.40M -Note that "filesystem" is really a dataset, so "pnfs-dataset" seems odd. Could it be made "pnfs-data"?> Having the above be the default "zfs list" output is a simple change > to make, but I can''t commit do doing it unless, of course, there is > buy in from the ZFS team. Does anyone have insight into any known > problems it may cause to add the "type" property to the default output > or why it was left out in the first place?It should cause no problems -- scripts should be using -o and -H -- but I suppose it might. But I don''t see why it''s necessary. It''s already hard to tell if a dataset is a filesystem or a zvol from the default output when you consider filesystems that are not mounted anywhere, thus pNFS wouldn''t be creating an ambiguity -- the ambiguity exists now and it''s not really a big deal. Nico --
On Mar 4, 2009, at 3:10 PM, Nicolas Williams wrote:> On Wed, Mar 04, 2009 at 02:13:51PM -0700, Lisa Week wrote: >> (pnfs-17-21:/home/lisagab):6 % zfs list -o >> name,type,used,avail,refer,mountpoint >> NAME TYPE USED AVAIL REFER >> MOUNTPOINT >> rpool filesystem 30.0G 37.0G 32.5K / >> rpool >> rpool/ROOT filesystem 18.2G 37.0G 18K >> legacy >> rpool/ROOT/snv_105 filesystem 18.2G 37.0G 6.86G / >> rpool/ROOT/snv_105/var filesystem 11.4G 37.0G 11.4G / >> var >> rpool/dump volume 9.77G 37.0G 9.77G - >> rpool/export filesystem 40K 37.0G 21K / >> export >> rpool/export/home filesystem 19K 37.0G 19K / >> export/home >> rpool/pnfsds pnfs-dataset 31K 37.0G 15K - >> rpool/pnfsds/47C80414080A4A42 pnfs-dataset 16K 37.0G 16K - >> rpool/swap volume 1.97G 38.9G 4.40M - > > Note that "filesystem" is really a dataset, so "pnfs-dataset" seems > odd. > Could it be made "pnfs-data"?Yes, I agree. "pnfs-dataset" is awkward.> > >> Having the above be the default "zfs list" output is a simple change >> to make, but I can''t commit do doing it unless, of course, there is >> buy in from the ZFS team. Does anyone have insight into any known >> problems it may cause to add the "type" property to the default >> output >> or why it was left out in the first place? > > It should cause no problems -- scripts should be using -o and -H -- > but > I suppose it might. But I don''t see why it''s necessary. It''s already > hard to tell if a dataset is a filesystem or a zvol from the default > output when you consider filesystems that are not mounted anywhere, > thus > pNFS wouldn''t be creating an ambiguity -- the ambiguity exists now and > it''s not really a big deal.My (humble) opinion is: Even though it is hard to tell if a dataset is a filesystem or a zvol now, it doesn''t mean we can''t make it better... Thanks, Lisa
On Wed, Mar 04, 2009 at 03:49:54PM -0700, Lisa Week wrote:> My (humble) opinion is: Even though it is hard to tell if a dataset is > a filesystem or a zvol now, it doesn''t mean we can''t make it better...Agreed.
On Wed, March 4, 2009 15:13, Lisa Week wrote:> Does anyone have insight into any known > problems it may cause to add the "type" property to the default output > or why it was left out in the first place?The obvious problem that always comes up in this situation is existing scripts that parse the output. I''m writing several myself, but using -H where available. And where -H is available, I think you should feel free to change the default output when -H is not present, on the grounds that scripts parsing output SHOULD use -H. In my own scripts, I often use -o to specify the output even if it''s default or very close to default, as defense against default output changing on me. -- David Dyer-Bennet, dd-b at dd-b.net; http://dd-b.net/ Snapshots: http://dd-b.net/dd-b/SnapshotAlbum/data/ Photos: http://dd-b.net/photography/gallery/ Dragaera: http://dragaera.info