Hi, Unfortunately, every now and then someone has his zpool corrupt, with no tools to fix it! This is due to either zfs bugs, or hardware lying about whether the bits really hit the platters. I am evaluating what I should be using for storing VMware ESX VM images (ext3 or zfs on NFS). I really really want zfs snapshots, but loosing the pool is going to be a royal pain for small businesses. My questions are: 1- What are the best practices to avoid pool corruption (even if it incurs a performance hit) ? 2- I remember a suggested idea that zfs would iterate back in time when mounting a zpool till it finds a fully written pool and uses that, thus avoiding corruption. Is there an RFE for that yet ? I''d like to subscribe to that, and I might even delay jumping on the zfs wagon till it''s got this recovery feature! Regards -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://mail.opensolaris.org/pipermail/zfs-discuss/attachments/20081018/f0904a25/attachment.html>
Well, I don''t have a huge amount of experience in Solaris, but I can certainly share my thoughts. 1. BACKUPS Always ensure you have backups of the pool. Ideally stored in a neutral format. Our plan is to ensure that all ZFS stores are also stored: - on tape via star - on an off-site ZFS system, synchronised with zfs send / receive. 2. CHOICE OF HARDWARE Research your hardware and drivers carefully, and test failure modes where possible. ZFS availability depends heavily on the behaviour of the hardware and drivers. My personal preference would be: - Sun hardware (Choose a platform like the x4500, x4540, x4240 designed for running Solaris & ZFS) - Trusted hardware with good Solaris support (eg. Areca raid controllers with manufacturer supplied drivers). - 3rd party hardware using the same chipsets Sun use (Supermicro AOC-SAT2-MV8 sata cards, or the LSI LSISAS3081E-R). While these are probably the recommended cards for ZFS, both have some caveats. The Supermicro one has some hot swap issues. The way SAS drives are labelled means that with the LSI one, in time it will get hard to know which disk is which for fault finding. Personally I wouldn''t want to run ZFS on anything outside of that list, and after seeing how well it integrates with the hardware on an x4500, I''m leaning heavily towards preferring it on Sun hardware. -- This message posted from opensolaris.org
Same situation here. We are just planning a new 200TB+ NFS storage pool and a bunch of x4100+j4400 would be great (j4400 is missing multipath??) and cheap but we already lost many zpools in the last 2 years and so ZFS won''t be an option. At this time we are using ZFS only in our central backup storage. To tell the truth we had no corruption starting from snv_60 but can''t consider seriously using ZFS without "zfschk" :( gino (will buy Netapp/HDS until zfschk comes out) -- This message posted from opensolaris.org