Hello all, I''ve tried changing all kinds of attributes for the zfs''s, but I can''t seem to find the right configuration. So I''m trying to move some zfs''s under another, it looks like this: /pool/joe_user move to /pool/homes/joe_user I know I can do this with zfs rename, and everything is fine. The problem I''m having is, when I mount /pool/homes I see joe_user directory but its empty. I know for a fact there are files that are in the directory. Am I on the right path, thinking its a permissions issue? If it matters, I''m using zfs version 4 but I having been trying to twist other knobs in version 14. Any help would be appreciated. Thanks, Gregory
Gregory Skelton wrote:> I''ve tried changing all kinds of attributes for the zfs''s, but I can''t > seem to find the right configuration. > > So I''m trying to move some zfs''s under another, it looks like this: > > /pool/joe_user move to /pool/homes/joe_userYou can do it in several ways: 1. Create a new FS and copy data from old FS :) 2. Change mountpoint: # zfs set mountpoint=/pool/homes/joe_user pool/joe_user 3. Use clone and promote: # zfs snapshot pool/joe_user at copy # zfs clone pool/joe_user at copy pool/homes/joe_user # zfs promote pool/homes/joe_user verify that all is ok, then destroy old FS It''s IMHO... -- WBR, Andrey V. Elsukov
Thanks for your reply Andrey, Unfortunately, I wasn''t able to see the files in that directory from any of scenario''s. It seems a zfs inside another zfs is causing the problem. Of course when I put just a directory inside the zfs, everything can be seen. Cheers, Greg On Fri, 21 Aug 2009, Andrey V. Elsukov wrote:> Gregory Skelton wrote: >> I''ve tried changing all kinds of attributes for the zfs''s, but I can''t >> seem to find the right configuration. >> >> So I''m trying to move some zfs''s under another, it looks like this: >> >> /pool/joe_user move to /pool/homes/joe_user > > You can do it in several ways: > 1. Create a new FS and copy data from old FS :) > 2. Change mountpoint: > # zfs set mountpoint=/pool/homes/joe_user pool/joe_user > 3. Use clone and promote: > # zfs snapshot pool/joe_user at copy > # zfs clone pool/joe_user at copy pool/homes/joe_user > # zfs promote pool/homes/joe_user > verify that all is ok, then destroy old FS > > It''s IMHO... > -- > WBR, Andrey V. Elsukov >-- Gregory R. Skelton Phone: (414)229-2678 (Office) System Administrator: (920) 246-4415 (Cell) 1900 E. Kenwood Blvd: gskelton at gravity.phys.uwm.edu University of Wisconsin : AIM/ICQ gregor159 Milwaukee, WI 53201 lsc-group.phys.uwm.edu/~gskelton Emergency Email: gregorsk at vzw.blackberry.net
if you are talking about NFS, this is due to how ZFS file systems work. When you share a ZFS filesystem via NFS it will share everything IN that filesystem but if you have 2 filesystems, it will only share that second fs''s mount point. what i mean is, if you have something like pool/filesystem and pool/filesystem/filesystem2 and you share pool/filesystem via NFS it will only share what is actually in that filesystem....so you will need to also share the second filesystem as well. On Fri, Aug 21, 2009 at 3:41 PM, Gregory Skelton < gskelton at gravity.phys.uwm.edu> wrote:> Thanks for your reply Andrey, > > Unfortunately, I wasn''t able to see the files in that directory from any of > scenario''s. It seems a zfs inside another zfs is causing the problem. Of > course when I put just a directory inside the zfs, everything can be seen. > > Cheers, > Greg > > > On Fri, 21 Aug 2009, Andrey V. Elsukov wrote: > > Gregory Skelton wrote: >> >>> I''ve tried changing all kinds of attributes for the zfs''s, but I can''t >>> seem to find the right configuration. >>> >>> So I''m trying to move some zfs''s under another, it looks like this: >>> >>> /pool/joe_user move to /pool/homes/joe_user >>> >> >> You can do it in several ways: >> 1. Create a new FS and copy data from old FS :) >> 2. Change mountpoint: >> # zfs set mountpoint=/pool/homes/joe_user pool/joe_user >> 3. Use clone and promote: >> # zfs snapshot pool/joe_user at copy >> # zfs clone pool/joe_user at copy pool/homes/joe_user >> # zfs promote pool/homes/joe_user >> verify that all is ok, then destroy old FS >> >> It''s IMHO... >> -- >> WBR, Andrey V. Elsukov >> >> > -- > Gregory R. Skelton Phone: (414)229-2678 (Office) > System Administrator: (920) 246-4415 (Cell) > 1900 E. Kenwood Blvd: gskelton at gravity.phys.uwm.edu > University of Wisconsin : AIM/ICQ gregor159 > Milwaukee, WI 53201 lsc-group.phys.uwm.edu/~gskelton<lsc-group.phys.uwm.edu/~gskelton> > Emergency Email: gregorsk at vzw.blackberry.net > > _______________________________________________ > zfs-discuss mailing list > zfs-discuss at opensolaris.org > mail.opensolaris.org/mailman/listinfo/zfs-discuss >-------------- next part -------------- An HTML attachment was scrubbed... URL: <mail.opensolaris.org/pipermail/zfs-discuss/attachments/20090822/28f05c83/attachment.html>
Hi Thomas, Yes, that is exactly whats happening to us. I''ve tried to "share" the zfs inside the other zfs. Like so, but I''m still seeing an empty directory. export1/dfazi sharenfs rw local So in our particular setup, we have the following: export1/dfazi 137G 2.05T 137G /export/home/dfazi We''ve mounted /export/home, and we''re trying to see the contents of dfazi but because its a different zfs we''re seeing nothing. Is there anything else I can do to fix this? Here are the properties of /export1/dfazi Thanks for all the help so far, Cheers, Gregory NAME PROPERTY VALUE SOURCE export1/dfazi type filesystem - export1/dfazi creation Mon Jul 27 13:04 2009 - export1/dfazi used 137G - export1/dfazi available 2.05T - export1/dfazi referenced 137G - export1/dfazi compressratio 1.00x - export1/dfazi mounted yes - export1/dfazi quota none default export1/dfazi reservation none default export1/dfazi recordsize 128K default export1/dfazi mountpoint /export/home/dfazi inherited from export1 export1/dfazi sharenfs rw local export1/dfazi checksum on default export1/dfazi compression off default export1/dfazi atime on default export1/dfazi devices on default export1/dfazi exec on default export1/dfazi setuid on default export1/dfazi readonly off default export1/dfazi zoned off default export1/dfazi snapdir hidden default export1/dfazi aclmode discard local export1/dfazi aclinherit discard local export1/dfazi canmount on default export1/dfazi shareiscsi on local export1/dfazi xattr off local export1/dfazi copies 1 default export1/dfazi version 1 - export1/dfazi utf8only off - export1/dfazi normalization none - export1/dfazi casesensitivity sensitive - export1/dfazi vscan off default export1/dfazi nbmand off default export1/dfazi sharesmb off default export1/dfazi refquota none default export1/dfazi refreservation none default export1/dfazi primarycache all default export1/dfazi secondarycache all default export1/dfazi usedbysnapshots 7.48M - export1/dfazi usedbydataset 137G - export1/dfazi usedbychildren 0 - export1/dfazi usedbyrefreservation 0 - On Sat, 22 Aug 2009, Thomas Burgess wrote:> if you are talking about NFS, this is due to how ZFS file systems work. > When you share a ZFS filesystem via NFS it will share everything IN that > filesystem but if you have 2 filesystems, it will only share that second > fs''s mount point. > > > what i mean is, if you have something like pool/filesystem > and pool/filesystem/filesystem2 > and you share pool/filesystem via NFS it will only share what is actually in > that filesystem....so you will need to also share the second filesystem as > well. > > On Fri, Aug 21, 2009 at 3:41 PM, Gregory Skelton < > gskelton at gravity.phys.uwm.edu> wrote: > >> Thanks for your reply Andrey, >> >> Unfortunately, I wasn''t able to see the files in that directory from any of >> scenario''s. It seems a zfs inside another zfs is causing the problem. Of >> course when I put just a directory inside the zfs, everything can be seen. >> >> Cheers, >> Greg >> >> >> On Fri, 21 Aug 2009, Andrey V. Elsukov wrote: >> >> Gregory Skelton wrote: >>> >>>> I''ve tried changing all kinds of attributes for the zfs''s, but I can''t >>>> seem to find the right configuration. >>>> >>>> So I''m trying to move some zfs''s under another, it looks like this: >>>> >>>> /pool/joe_user move to /pool/homes/joe_user >>>> >>> >>> You can do it in several ways: >>> 1. Create a new FS and copy data from old FS :) >>> 2. Change mountpoint: >>> # zfs set mountpoint=/pool/homes/joe_user pool/joe_user >>> 3. Use clone and promote: >>> # zfs snapshot pool/joe_user at copy >>> # zfs clone pool/joe_user at copy pool/homes/joe_user >>> # zfs promote pool/homes/joe_user >>> verify that all is ok, then destroy old FS >>> >>> It''s IMHO... >>> -- >>> WBR, Andrey V. Elsukov >>> >>> >> -- >> _______________________________________________ >> zfs-discuss mailing list >> zfs-discuss at opensolaris.org >> mail.opensolaris.org/mailman/listinfo/zfs-discuss >> >
you could mount both, but you should talk to the nfs/zfs experts about the proper way. Theres more than one method of doing so. What i did was something like this: i had a ZFS share called /store and another at /store/Video i wanted to share them via nfs, so i mount /store on /mnt/store on my local system and i mounted /store/Video on /mnt/store/Video but i could have just as easily mounted it somewhere else. On Mon, Aug 24, 2009 at 12:00 PM, Gregory Skelton < gskelton at gravity.phys.uwm.edu> wrote:> Hi Thomas, > > Yes, that is exactly whats happening to us. I''ve tried to "share" the zfs > inside the other zfs. Like so, but I''m still seeing an empty directory. > > export1/dfazi sharenfs rw local > > So in our particular setup, we have the following: > export1/dfazi 137G 2.05T 137G /export/home/dfazi > > We''ve mounted /export/home, and we''re trying to see the contents of dfazi > but because its a different zfs we''re seeing nothing. Is there anything else > I can do to fix this? Here are the properties of /export1/dfazi > > Thanks for all the help so far, > Cheers, > Gregory > > > NAME PROPERTY VALUE SOURCE > export1/dfazi type filesystem - > export1/dfazi creation Mon Jul 27 13:04 2009 - > export1/dfazi used 137G - > export1/dfazi available 2.05T - > export1/dfazi referenced 137G - > export1/dfazi compressratio 1.00x - > export1/dfazi mounted yes - > export1/dfazi quota none default > export1/dfazi reservation none default > export1/dfazi recordsize 128K default > export1/dfazi mountpoint /export/home/dfazi inherited from > export1 > export1/dfazi sharenfs rw local > export1/dfazi checksum on default > export1/dfazi compression off default > export1/dfazi atime on default > export1/dfazi devices on default > export1/dfazi exec on default > export1/dfazi setuid on default > export1/dfazi readonly off default > export1/dfazi zoned off default > export1/dfazi snapdir hidden default > export1/dfazi aclmode discard local > export1/dfazi aclinherit discard local > export1/dfazi canmount on default > export1/dfazi shareiscsi on local > export1/dfazi xattr off local > export1/dfazi copies 1 default > export1/dfazi version 1 - > export1/dfazi utf8only off - > export1/dfazi normalization none - > export1/dfazi casesensitivity sensitive - > export1/dfazi vscan off default > export1/dfazi nbmand off default > export1/dfazi sharesmb off default > export1/dfazi refquota none default > export1/dfazi refreservation none default > export1/dfazi primarycache all default > export1/dfazi secondarycache all default > export1/dfazi usedbysnapshots 7.48M - > export1/dfazi usedbydataset 137G - > export1/dfazi usedbychildren 0 - > export1/dfazi usedbyrefreservation 0 - > > > > > On Sat, 22 Aug 2009, Thomas Burgess wrote: > > if you are talking about NFS, this is due to how ZFS file systems work. >> When you share a ZFS filesystem via NFS it will share everything IN that >> filesystem but if you have 2 filesystems, it will only share that second >> fs''s mount point. >> >> >> what i mean is, if you have something like pool/filesystem >> and pool/filesystem/filesystem2 >> and you share pool/filesystem via NFS it will only share what is actually >> in >> that filesystem....so you will need to also share the second filesystem as >> well. >> >> On Fri, Aug 21, 2009 at 3:41 PM, Gregory Skelton < >> gskelton at gravity.phys.uwm.edu> wrote: >> >> Thanks for your reply Andrey, >>> >>> Unfortunately, I wasn''t able to see the files in that directory from any >>> of >>> scenario''s. It seems a zfs inside another zfs is causing the problem. Of >>> course when I put just a directory inside the zfs, everything can be >>> seen. >>> >>> Cheers, >>> Greg >>> >>> >>> On Fri, 21 Aug 2009, Andrey V. Elsukov wrote: >>> >>> Gregory Skelton wrote: >>> >>>> >>>> I''ve tried changing all kinds of attributes for the zfs''s, but I can''t >>>>> seem to find the right configuration. >>>>> >>>>> So I''m trying to move some zfs''s under another, it looks like this: >>>>> >>>>> /pool/joe_user move to /pool/homes/joe_user >>>>> >>>>> >>>> You can do it in several ways: >>>> 1. Create a new FS and copy data from old FS :) >>>> 2. Change mountpoint: >>>> # zfs set mountpoint=/pool/homes/joe_user pool/joe_user >>>> 3. Use clone and promote: >>>> # zfs snapshot pool/joe_user at copy >>>> # zfs clone pool/joe_user at copy pool/homes/joe_user >>>> # zfs promote pool/homes/joe_user >>>> verify that all is ok, then destroy old FS >>>> >>>> It''s IMHO... >>>> -- >>>> WBR, Andrey V. Elsukov >>>> >>>> >>>> -- >>> _______________________________________________ >>> zfs-discuss mailing list >>> zfs-discuss at opensolaris.org >>> mail.opensolaris.org/mailman/listinfo/zfs-discuss >>> >>> >>-------------- next part -------------- An HTML attachment was scrubbed... URL: <mail.opensolaris.org/pipermail/zfs-discuss/attachments/20090824/e122c072/attachment.html>
Gregory Skelton wrote:> Yes, that is exactly whats happening to us. I''ve tried to "share" the > zfs inside the other zfs. Like so, but I''m still seeing an empty directory.What are you using for a client? What version of NFS? NFSv4 in Solaris Nevada build 77 and later, or any OpenSolaris versions, will do this "mirror mounts" stuff automatically. Other than that, it''s manual mounts or automounter map entries. Rob T
Hi,> What are you using for a client? What version of NFS?We''re using Red Hat Enterprise Linux(Centos) 5.3 for the clients, with nfs 3> NFSv4 in Solaris Nevada build 77 and later, or any OpenSolaris > versions, will do this "mirror mounts" stuff automatically. > Other than that, it''s manual mounts or automounter map entries.The mirrored mount sound interesting, and looking at it; that''s exactly what we want, but its starting to seem like its impossible. I think the "mirrored mount" is what we''re seeing, and thats why we can see the directories, which is great but working with the files inside is another story. If its possible, I''d still like to mount the base dir, and get the sub directories (other zfs''s) access for free. Thanks for all the help, Gregory> > Rob T > >
Gregory Skelton wrote:>> What are you using for a client? What version of NFS? > > We''re using Red Hat Enterprise Linux(Centos) 5.3 for the clients, with > nfs 3You should try NFSv4 - Linux NFSv4 support came in with this "mirror mount" support.> If its possible, I''d still like to mount the base dir, and get the sub > directories (other zfs''s) access for free.Some other OSs have code to make their servers pretend nested mounts are part of the same big filesystem, but we don''t like the issues that that causes, and won''t add such support to the Solaris server. Rob T
Thanks Robert, The older version of nfs that was exactly the problem. Nfs3 made the directorieslook like they were empty, while nfs4 displays the contents of each directory. Many Thanks! Gregory On Mon, 24 Aug 2009, Robert Thurlow wrote:> Gregory Skelton wrote: > >> > What are you using for a client? What version of NFS? >> >> We''re using Red Hat Enterprise Linux(Centos) 5.3 for the clients, with nfs >> 3 > > You should try NFSv4 - Linux NFSv4 support came in with this > "mirror mount" support. > >> If its possible, I''d still like to mount the base dir, and get the sub >> directories (other zfs''s) access for free. > > Some other OSs have code to make their servers pretend nested > mounts are part of the same big filesystem, but we don''t like > the issues that that causes, and won''t add such support to the > Solaris server. > > Rob T >