On 12 April 2011 15:39, Greg Earle <[email protected]> wrote: > On Apr 12, 2011, at 5:16 AM, Ben Taylor wrote: > >> In LDOM 2.0, you can export single slices. >> >> However, I'm wondering why the OP isn't just exporting the 2 whole disks >> and doing the SVM and ZFS work locally to the LDOM. >> >> It seems much more complicate exporting a bunch of single slices to >> make it look like there are physical disks, and expecting the master ldom >> to manage the SVM config. Seems to me that this could introduce a >> unintended consequence of having to managed the volume from two >> different ldoms. > > You can export single slices in LDom 1.3 too, that's part of what makes no > sense to me - we have another, older T-5140 running LDom 1.3 and Solaris > 10/09 in the control LDom (10/08 in the client, both cases) that is doing > nothing *but* exporting slices, and having the ZFS zpool created from 2 > exported slices works just fine.
Yep, sorry for misunderstanding. But if you export it as a slice it stays a slice. So if you try instead of "zpool create -f dbdump mirror c0d2s0 c0d3s0" zpool create -f dbdump mirror c0d2s1 c0d3s1 nothing changes? -Petr > > Primary LDom on the older T-5140: > > DISK > NAME VOLUME TOUT DEVICE SERVER > MPGROUP > vdisk0 c1t0d0s3@primary-vds0 disk@0 primary > vdisk1 c1t1d0s3@primary-vds0 disk@1 primary > cdrom dvd_vol@primary-vds0 disk@2 primary > vdisk3 c1t2d0s0@primary-vds0 disk@3 primary > vdisk4 c1t3d0s0@primary-vds0 disk@4 primary > vdisk5 c1t3d0s3@primary-vds0 disk@5 primary > vdisk6 c1t2d0s3@primary-vds0 disk@6 primary > vdisk7 c1t2d0s4@primary-vds0 disk@7 primary > vdisk8 c1t3d0s4@primary-vds0 disk@8 primary > vdisk9 c1t2d0s1@primary-vds0 disk@9 primary <- ZFS > slice > vdisk10 c1t3d0s1@primary-vds0 disk@10 primary <- ZFS > slice > vdisk11 c1t2d0s7@primary-vds0 disk@11 primary > vdisk12 c1t3d0s7@primary-vds0 disk@12 primary > > 8db:1:1008 [/] # format > Searching for disks...done > > > AVAILABLE DISK SELECTIONS: > 0. c0d0 <SUN-DiskSlice-20GB cyl 2125 alt 2 hd 24 sec 848> > /virtual-devices@100/channel-devices@200/disk@0 > 1. c0d1 <SUN-DiskSlice-20GB cyl 2125 alt 2 hd 24 sec 848> > /virtual-devices@100/channel-devices@200/disk@1 > 2. c0d3 <SUN-DiskSlice-100GB cyl 16779 alt 2 hd 20 sec 625> > /virtual-devices@100/channel-devices@200/disk@3 > 3. c0d4 <SUN-DiskSlice-100GB cyl 16779 alt 2 hd 20 sec 625> > /virtual-devices@100/channel-devices@200/disk@4 > 4. c0d5 <SUN-DiskSlice-50GB cyl 8390 alt 2 hd 20 sec 625> > /virtual-devices@100/channel-devices@200/disk@5 > 5. c0d6 <SUN-DiskSlice-50GB cyl 8390 alt 2 hd 20 sec 625> > /virtual-devices@100/channel-devices@200/disk@6 > 6. c0d7 <SUN-DiskSlice-29GB cyl 4867 alt 2 hd 20 sec 625> > /virtual-devices@100/channel-devices@200/disk@7 > 7. c0d8 <SUN-DiskSlice-29GB cyl 4867 alt 2 hd 20 sec 625> > /virtual-devices@100/channel-devices@200/disk@8 > 8. c0d9 <SUN-DiskSlice-100GB cyl 16779 alt 2 hd 20 sec 625> > /virtual-devices@100/channel-devices@200/disk@9 <- For > ZFS > 9. c0d10 <SUN-DiskSlice-100GB cyl 16779 alt 2 hd 20 sec 625> > /virtual-devices@100/channel-devices@200/disk@a <- For > ZFS > 10. c0d11 <SUN-DiskSlice-378MB cyl 63 alt 2 hd 20 sec 625> > /virtual-devices@100/channel-devices@200/disk@b > 11. c0d12 <SUN-DiskSlice-378MB cyl 63 alt 2 hd 20 sec 625> > /virtual-devices@100/channel-devices@200/disk@c > > Specify disk (enter its number): 9 > selecting c0d10 > [disk formatted, no defect list found] > /dev/dsk/c0d10s0 is part of active ZFS pool dbdump. Please see zpool(1M). > > mipl8db:1:1009 [/] # zpool status > pool: dbdump > state: ONLINE > scrub: none requested > config: > > NAME STATE READ WRITE CKSUM > dbdump ONLINE 0 0 0 > mirror ONLINE 0 0 0 > c0d9s0 ONLINE 0 0 0 > c0d10s0 ONLINE 0 0 0 > > errors: No known data errors > > pool: rpool > state: ONLINE > scrub: none requested > config: > > NAME STATE READ WRITE CKSUM > rpool ONLINE 0 0 0 > mirror ONLINE 0 0 0 > c0d0s0 ONLINE 0 0 0 > c0d1s0 ONLINE 0 0 0 > > errors: No known data errors > > Having 2 sets of slices with ZFS zpools created on them works perfectly fine > there; but on the newer T-5120 with Oracle VM Server 2.0 and Solaris 9/10 in > the control (10/08 in the client) LDom, I get the aforementioned I/O error > when I try to create the zpool using essentially the same exact command line > as on this T-5140 (modulo slice numbers): > > db-dev:1:76 [/] # zpool create -f dbdump mirror c0d2s0 c0d3s0 > cannot create 'dbdump': one or more devices is currently unavailable > db-dev:1:79 [/] # zpool create -f dbdump c0d2s0 > cannot create 'dbdump': I/O error > db-dev:1:80 [/] # zpool create -f dbdump c0d3s0 > cannot create 'dbdump': I/O error > > As mentioned before, the 2 slices in the control LDom are no different than > the ones from the T-5140 setup that works: > > DISK > NAME VOLUME TOUT ID DEVICE SERVER > MPGROUP > vdisk0 db-dev-vol0@primary-vds0 0 disk@0 primary > [...] > vdisk1 c1t2d0s1@primary-vds0 2 disk@2 primary <- > ZFS > vdisk2 c1t3d0s1@primary-vds0 3 disk@3 primary <- > ZFS > > db-dev:1:81 [/] # format > Searching for disks...done > > > AVAILABLE DISK SELECTIONS: > 0. c0d0 <SUN-DiskImage-20GB cyl 566 alt 2 hd 96 sec 768> > /virtual-devices@100/channel-devices@200/disk@0 > 1. c0d2 <SUN-DiskSlice-100GB cyl 16780 alt 2 hd 20 sec 625> > /virtual-devices@100/channel-devices@200/disk@2 <- For > ZFS > 2. c0d3 <SUN-DiskSlice-100GB cyl 16780 alt 2 hd 20 sec 625> > /virtual-devices@100/channel-devices@200/disk@3 <- For > ZFS > 3. c0d4 <SUN-DiskImage-99GB cyl 2842 alt 2 hd 96 sec 768> > /virtual-devices@100/channel-devices@200/disk@4 > 4. c0d5 <SUN-DiskImage-50GB cyl 1420 alt 2 hd 96 sec 768> > /virtual-devices@100/channel-devices@200/disk@5 > 5. c0d6 <SUN-DiskImage-29GB cyl 822 alt 2 hd 96 sec 768> > /virtual-devices@100/channel-devices@200/disk@6 > I am completely baffled. > > I also see no reason why I can't export the SVM metadevices from the > back-end. Like Petr said, zvols are supposed to work and they aren't much > different than SVM metadevice mirrors conceptually. > > Ben, to answer your question, this is a development database server; the > final config will probably not be what I am currently trying out size-wise > which is the same sizes as the operational database server (they will end up > being much smaller than the current test partitions). > > In other words, we don't want to export the entire 2 disks to the database > client LDom, as we might want to reserve a couple of slices for other > potential client LDoms in the same host. If we *have* to (because we can't > get this to work), we will - we'd just rather be able to export a certain > percentage of the 2 disks, but not all of them. > > The reason I am trying to do the SVM setup on the back-end in this newer > machine was that I reasoned that doing the SVM mirroring and I/O closer to > the hardware was a better idea than exporting 'dumb' slices and doing the SVM > metadevice setup in the client LDom. On the ZFS side it is just to hold the > database backups so speed is not really an issue and thus that is why we are > doing the ZFS setup on the client LDom end in both cases. I can of course > try creating a zvol out of the 2 slices on the control LDom end and export > that instead. It may come to that ... > > Thanks, > > - Greg > > _______________________________________________ > ldoms-discuss mailing list > [email protected] > http://mail.opensolaris.org/mailman/listinfo/ldoms-discuss > _______________________________________________ ldoms-discuss mailing list [email protected] http://mail.opensolaris.org/mailman/listinfo/ldoms-discuss
