I will be setting up a NexentaStor Community Edition based ZFS file server. I will be serving some zvols over iSCSI to some FreeBSD machines to host jails in. 1) The ZFS box offers a single iSCSI target that exposes all the zvols as individual disks. When the FreeBSD initiator finds it, it creates a separate disk for each zvol. I assume if I have multiple FreeBSD machines connecting to this iSCSI target, as long as no individual zvol is mounted on more than 1 FreeBSD machine, the fact that a disk "exists" for each zvol on each FreeBSD machine is irrelevant and won''t cause problems 2) I am thinking about formatting the virtual disks served from the Nexenta iSCSI target as ZFS on the FreeBSD machine even though it has no redundancy. I see this as safe since the backing store on the Nexenta machine is a redundant based ZFS zvol... Is this correct thinking? Thanks Chad
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA256 On 11/11/10 17:57, Chad Leigh -- Shire.Net LLC wrote:> > I will be setting up a NexentaStor Community Edition based ZFS file > server. I will be serving some zvols over iSCSI to some FreeBSD > machines to host jails in. > > 1) The ZFS box offers a single iSCSI target that exposes all the > zvols as individual disks. When the FreeBSD initiator finds it, it > creates a separate disk for each zvol. I assume if I have multiple > FreeBSD machines connecting to this iSCSI target, as long as no > individual zvol is mounted on more than 1 FreeBSD machine, the fact > that a disk "exists" for each zvol on each FreeBSD machine is > irrelevant and won''t cause problemsThis is correct.> 2) I am thinking about formatting the virtual disks served from the > Nexenta iSCSI target as ZFS on the FreeBSD machine even though it has > no redundancy. I see this as safe since the backing store on the > Nexenta machine is a redundant based ZFS zvol... Is this correct > thinking?Speaking for redundancy, I don''t think there is much point adding redundancy at client side when you already have at the iSCSI target. However, it makes me to feel a bit nervous since the added transmission chains (iSCSI stack, TCP/IP stack, network driver, network card, ethernet cable, switch, then vice versa) could increase the risk of introducing data errors during transmission (note that iSCSI uses TCP which do have built in checksuming so the likelihood is very low, though). Cheers, - -- Xin LI <delphij at delphij.net> http://www.delphij.net/ FreeBSD - The Power to Serve! Live free or die -----BEGIN PGP SIGNATURE----- Version: GnuPG v2.0.16 (FreeBSD) iQEcBAEBCAAGBQJM3KP/AAoJEATO+BI/yjfB2uEIAK8zADbsEpyRelgRBJnelBT+ P4EbBIfexWxy29RXoqWwcXi7tOr1UqpaeqzgqfhJmz0kX42by4CYS6rPIffaaZVq yUZ8TItUXo1v2TK3k/S3XBlEdDO4A4bHrVgv/+ylqx+VyNPun7wtbHgKgb2ZANbr +PWevIm4QOPHDApnznbeigTSCqVwIC31G2M8pFseDZufV/aKWnYYG4G1AtXi1vDq ABELVQz23HgFTaFnNqrFrCYKU0oGi2ff2TDE84t2apoqxvfp+esbDxBI2BKt14DQ m14lBdPvxE0nEWcPtm1ica5Lc45GZKPYHyNrL1h6KeAq8CNQC90CKYMMZ0+mtsA=VvvJ -----END PGP SIGNATURE-----
On Nov 11, 2010, at 7:18 PM, Xin LI wrote:> -----BEGIN PGP SIGNED MESSAGE----- > Hash: SHA256 > > On 11/11/10 17:57, Chad Leigh -- Shire.Net LLC wrote: >> >> I will be setting up a NexentaStor Community Edition based ZFS file >> server. I will be serving some zvols over iSCSI to some FreeBSD >> machines to host jails in. >> >> 1) The ZFS box offers a single iSCSI target that exposes all the >> zvols as individual disks. When the FreeBSD initiator finds it, it >> creates a separate disk for each zvol. I assume if I have multiple >> FreeBSD machines connecting to this iSCSI target, as long as no >> individual zvol is mounted on more than 1 FreeBSD machine, the fact >> that a disk "exists" for each zvol on each FreeBSD machine is >> irrelevant and won''t cause problems > > This is correct.A follow-on question. If the zvol (virtual disks) are mounted READ ONLY, is it possible to mount it on multiple FreeBSD systems at the same time and access it for reading only from all the systems? (With only one system having it R/W and that only being used occasionally when the new software needs to be installed for the jails)? What I want to do does not rely on this but could make things easier for me...
> From: zfs-discuss-bounces at opensolaris.org [mailto:zfs-discuss- > bounces at opensolaris.org] On Behalf Of Chad Leigh -- Shire.Net LLC > > >> 1) The ZFS box offers a single iSCSI target that exposes all the > >> zvols as individual disks. When the FreeBSD initiator finds it, it > >> creates a separate disk for each zvol. I assume if I have multiple > >> FreeBSD machines connecting to this iSCSI target, as long as no > >> individual zvol is mounted on more than 1 FreeBSD machine, the fact > >> that a disk "exists" for each zvol on each FreeBSD machine is > >> irrelevant and won''t cause problems > > > > This is correct. > > A follow-on question. > > If the zvol (virtual disks) are mounted READ ONLY, is it possible to mountit on> multiple FreeBSD systems at the same time and access it for reading only > from all the systems? (With only one system having it R/W and that only > being used occasionally when the new software needs to be installed forthe> jails)? What I want to do does not rely on this but could make thingseasier> for me...Why are you sharing iscsi from nexenta to freebsd? Wouldn''t it be better for nexenta to simply create zfs filesystems, and then share nfs? Much more flexible in a lot of ways. Unless your design requirements require limiting the flexibility intentionally... I can''t think of any reason you''d want to do the iscsi thing from nexenta to freebsd.
On Nov 12, 2010, at 5:54 AM, Edward Ned Harvey wrote:> > Why are you sharing iscsi from nexenta to freebsd? Wouldn''t it be better > for nexenta to simply create zfs filesystems, and then share nfs? Much more > flexible in a lot of ways. Unless your design requirements require limiting > the flexibility intentionally... I can''t think of any reason you''d want to > do the iscsi thing from nexenta to freebsd. >Because for running jails (in very simple terms a FreeBSD jail is a really fancy chroot or a really simple approximation of a zone) it does not work very well (at least not in the past -- I have tried that but not recently). Things like apache don''t want to run off NFS mounted file system for example (actual httpd daemon -- not the webroots etc). -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://mail.opensolaris.org/pipermail/zfs-discuss/attachments/20101112/50a351da/attachment.html>