I'm in the process of trying to migrate my home file server from Solaris 10 (zpool v. 15, zfs v. 4) to FreeBSD 8.2. Hardware consists of SATA disks attached to both the motherboard (Gigabyte w/Intel CPU) and two internal Silicon Image-based SATA controllers. Zpools: Solaris root mirror (rpool), FreeBSD root mirror (zroot), and two pools in raidz of 3x 1TB (lake and lagoon).
I've kept my Solaris root pool disks unaltered (currently disconnected), and installed FreeBSD on an new root mirror.
I properly exported the zpools from Solaris 10 before attempting to import them into FreeBSD.
After booting into FreeBSD, I could not import both zpools. I was able to import lake, but it's disk devices did not show up in 'zpool status' but rather GUIDs. 'gpart show' seemed to indicate that FreeBSD thinks that the disks in lagoon are MBR disks. It also showed the GPT as being corrupted on the disks in lake. zdb -l dumped out the labels from a disk in the lake pool with no issues, though.
I've booted into Solaris 10 via DVD, re-imported both raidz pools, and their status is normal. I did notice that lake's disks show the devices in use as "c1t1d0" vs lagoon's "c2t5d0s0". Lagoon was originally created under an earlier version of Solaris 10, and zpool/zfs upgraded. zdb -l on lake also shows different GUIDs from those shown in FreeBSD.
I read in another thread from sub_mesa, where he had GPT issues in 8.2 that were recovered in 9.0 under ZFS v28.
Before I proceed any further, what should I be looking at to troubleshoot the issue on 8.2? Or should I switch to 9 and try the import there? Is 9's ZFS stable enough for general use? This system sees a lot of disk activity, mostly reads, however when writes happen they tend to be continuous for a few hours at a time. While much of the data on the is reproducible with work (DVD, BRD, and CD rips), this system is also a Time Machine target for the household Macs, so I'm concerned about integrity of personal data.
My other option is to go forward with upgrading the pool disks (planning to add 2x 2TB in a mirrored zpool then convert the 1TB disks to mirrored pools after data is migrated and the 2TB disks are burned in for a month or two). I'd do the GPT label and pool creation in FreeBSD to avoid having this issue crop up again, then do the data migration in Solaris.
This is my first serious foray into FreeBSD in many years, but I am a Solaris admin in my day job, so whatever is needed I can dig out.
I appreciate any suggestions and assistance.
I've kept my Solaris root pool disks unaltered (currently disconnected), and installed FreeBSD on an new root mirror.
I properly exported the zpools from Solaris 10 before attempting to import them into FreeBSD.
After booting into FreeBSD, I could not import both zpools. I was able to import lake, but it's disk devices did not show up in 'zpool status' but rather GUIDs. 'gpart show' seemed to indicate that FreeBSD thinks that the disks in lagoon are MBR disks. It also showed the GPT as being corrupted on the disks in lake. zdb -l dumped out the labels from a disk in the lake pool with no issues, though.
I've booted into Solaris 10 via DVD, re-imported both raidz pools, and their status is normal. I did notice that lake's disks show the devices in use as "c1t1d0" vs lagoon's "c2t5d0s0". Lagoon was originally created under an earlier version of Solaris 10, and zpool/zfs upgraded. zdb -l on lake also shows different GUIDs from those shown in FreeBSD.
I read in another thread from sub_mesa, where he had GPT issues in 8.2 that were recovered in 9.0 under ZFS v28.
Before I proceed any further, what should I be looking at to troubleshoot the issue on 8.2? Or should I switch to 9 and try the import there? Is 9's ZFS stable enough for general use? This system sees a lot of disk activity, mostly reads, however when writes happen they tend to be continuous for a few hours at a time. While much of the data on the is reproducible with work (DVD, BRD, and CD rips), this system is also a Time Machine target for the household Macs, so I'm concerned about integrity of personal data.
My other option is to go forward with upgrading the pool disks (planning to add 2x 2TB in a mirrored zpool then convert the 1TB disks to mirrored pools after data is migrated and the 2TB disks are burned in for a month or two). I'd do the GPT label and pool creation in FreeBSD to avoid having this issue crop up again, then do the data migration in Solaris.
This is my first serious foray into FreeBSD in many years, but I am a Solaris admin in my day job, so whatever is needed I can dig out.
I appreciate any suggestions and assistance.