I found this article (http://www.cuddletech.com/blog/pivot/entry.php?id=729) but I have 2 questions. I am trying the steps on Opensolaris build 54. Since you create the filesystem with newfs, isn''t that really a ufs filesystem running on top of zfs? Also I haven''t been able to do anything in the normal fashion (ie zfs and zpool commands) with the thin provisioned created filesytem. Can''t even mount it or online it. Was this just a demo of things to come or is thin provision ready for testing in zfs? This message posted from opensolaris.org
> I found this article > (http://www.cuddletech.com/blog/pivot/entry.php?id=729) but I have 2 > questions. I am trying the steps on Opensolaris build 54.> Since you create the filesystem with newfs, isn''t that really a ufs > filesystem running on top of zfs?In this case, yes. I wonder if you could create a second zfs pool on the volume. (Starting such pools at boot time might be problematic though!). The idea is that you have sparse raw storage available to you. The example placed a UFS filesystem on it, but you could do otherwise.> Also I haven''t been able to do > anything in the normal fashion (ie zfs and zpool commands) with the > thin provisioned created filesytem. Can''t even mount it or online it.No. It''s not a filesystem. It''s a zvol or raw volume of blocks.> Was this just a demo of things to come or is thin provision ready for > testing in zfs?It''s in there. Did you create the volumes as it shows (with the -s)? -- 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. >
> In this case, yes. I wonder if you could create a second zfs pool on > the volume. (Starting such pools at boot time might be problematic > though!). The idea is that you have sparse raw storage available to > you. The example placed a UFS filesystem on it, but you could do > otherwise.Followup to myself. Okay, it''s just wrong, but it appears to work.... :-) # zfs list NAME USED AVAIL REFER MOUNTPOINT tank 1.40M 8.24G 25.5K /tank tank/test 24.5K 8.24G 24.5K /tank/test tank/zvol1 1.28M 8.24G 1.28M - vtank 76K 1.91T 24.5K /vtank vtank is a zpool on top of tank/zvol1 NAME STATE READ WRITE CKSUM vtank ONLINE 0 0 0 /dev/zvol/dsk/tank/zvol1 ONLINE 0 0 0 # cp /usr/sbin/xntpdc /vtank # zfs list vtank NAME USED AVAIL REFER MOUNTPOINT vtank 178K 1.91T 126K /vtank I don''t think I want to reboot it in this state, though. :-) -- 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! I got led down the wrong path by following newfs. Now my other question is. How would you add raw storage to the vtank (virtual filesystem) as the usage approached the current underlying raw storage? Would you going forward just simply in the normal fashion ( i will try this when I can add another disk) zpool add vtank cXtXdX Is there a performance hit for having what seems to be a zfs on top a zpool on top a zpool? This message posted from opensolaris.org
Wade.Stuart at fallon.com
2007-Feb-02 16:49 UTC
[zfs-discuss] Re: ZFS and thin provisioning
zfs-discuss-bounces at opensolaris.org wrote on 02/02/2007 10:34:22 AM:> thanks Darren! I got led down the wrong path by following newfs. > > Now my other question is. How would you add raw storage to the vtank > (virtual filesystem) as the usage approached the current underlying > raw storage? > > Would you going forward just simply in the normal fashion ( i will > try this when I can add another disk) > zpool add vtank cXtXdX > > Is there a performance hit for having what seems to be a zfs on top > a zpool on top a zpool? >I would think so. Also a good test would be to write on the final fs a few blocks more data then the backing sparse volume actually has available. Gut feeling is that will cause a panic on current solaris. Is there any reason why this multi layering is even allowed -- seems risky and hackish?
> thanks Darren! I got led down the wrong path by following newfs. > > Now my other question is. How would you add raw storage to the vtank (virtual filesystem) as the usage approached the current underlying raw storage?You just increase the storage in the underlying pool. In my case, I''d just add storage to ''tank''.> Would you going forward just simply in the normal fashion ( i will try this when I can add another disk) > zpool add vtank cXtXdXNo. vtank is already large (but sparse). It''s ''tank'' that is limited and needs to be increased.> Is there a performance hit for having what seems to be a zfs on top a > zpool on top a zpool?I''m sure. Possibly significant. I did it based on Ben''s article just as an exercise. -- 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. >
> > Is there a performance hit for having what seems to be a zfs on top > > a zpool on top a zpool? > > I would think so. Also a good test would be to write on the final fs a few > blocks more data then the backing sparse volume actually has available. > Gut feeling is that will cause a panic on current solaris.That''s what I''m assuming. I might go ahead and force it later if I have time to watch the action. Of course I''ve done this on a somewhat old build. I''d rather do it on a recent one, and I''m unlikely to get around to that anytime soon on this machine.> Is there any reason why this multi layering is even allowed -- seems > risky and hackish?Is this something that an administrator is likely to do accidentally or without understanding the consequences? Even though risky, is really useless in all situations? Otherwise, is it worth code to prevent it? -- 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. >