Date: Mon, 7 Mar 2016 12:55:46 -0800 From: Liam Slusser <lslusser@gmail.com> To: zfs@lists.illumos.org Cc: "smartos-discuss@lists.smartos.org" <smartos-discuss@lists.smartos.org>, developer@lists.open-zfs.org, developer <developer@open-zfs.org>, illumos-developer <developer@lists.illumos.org>, omnios-discuss <omnios-discuss@lists.omniti.com>, Discussion list for OpenIndiana <openindiana-discuss@openindiana.org>, "zfs-discuss@list.zfsonlinux.org" <zfs-discuss@list.zfsonlinux.org>, "freebsd-fs@FreeBSD.org" <freebsd-fs@freebsd.org>, "zfs-devel@freebsd.org" <zfs-devel@freebsd.org> Subject: Re: [zfs] [developer] Re: [smartos-discuss] an interesting survey -- the zpool with most disks you have ever built Message-ID: <CAESZ%2B_-%2B1jKQC880bew-maDyZ_xnMmB7QxPHyKAc_3P44%2Bm%2BuQ@mail.gmail.com> In-Reply-To: <CALi05Xw1NGqZhXcS4HweX7AK0DU_mm01tj=rjB%2BqOU9N0-N=ng@mail.gmail.com> References: <95563acb-d27b-4d4b-b8f3-afeb87a3d599@me.com> <CACTb9pxJqk__DPN_pDy4xPvd6ETZtbF9y=B8U7RaeGnn0tKAVQ@mail.gmail.com> <CAJjvXiH9Wh%2BYKngTvv0XG1HtikWggBDwjr_MCb8=Rf276DZO-Q@mail.gmail.com> <56D87784.4090103@broken.net> <A5A6EA4AE9DCC44F8E7FCB4D6317B1D203178F1DD392@SH-MAIL.ISSI.COM> <5158F354-9636-4031-9536-E99450F312B3@RichardElling.com> <CALi05Xxm9Sdx9dXCU4C8YhUTZOwPY%2BNQqzmMEn5d0iFeOES6gw@mail.gmail.com> <6E2B77D1-E0CA-4901-A6BD-6A22C07536B3@gmail.com> <CALi05Xw1NGqZhXcS4HweX7AK0DU_mm01tj=rjB%2BqOU9N0-N=ng@mail.gmail.com>
next in thread | previous in thread | raw e-mail | index | archive | help
I don't have a 2000 drive array (thats amazing!) but I do have two 280 drive arrays which are in production. Here are the generic stats: server setup: OpenIndiana oi_151 1 server rack Dell r720xd 64g ram with mirrored 250g boot disks 5 x LSI 9207-8e dualport SAS pci-e host bus adapters Intel 10g fibre ethernet (dual port) 2 x SSD for log cache 2 x SSD for cache 23 x Dell MD1200 with 3T,4T, or 6T NLSAS disks (a mix of Toshiba, Western Digital, and Seagate drives - basically whatever Dell sends) zpool setup: 23 x 12-disk raidz2 glued together. 276 total disks. Basically each new 12 disk MD1200 is a new raidz2 added to the pool. Total size: ~797T We have an identical server which we replicate changes via zfs snapshots every few minutes. The whole setup as been up and running for a few years now, no issues. As we run low on space we purchase two additional MD1200 shelfs (one for each system) and add the new raidz2 into pool on-the-fly. The only real issues we've had is sometimes a disk fails in such a way (think Monty Python and the holy grail i'm not dead yet) where the disk hasn't failed but is timing out and slows the whole array to a standstill until we can manual find and remove the disk. Other problems are once a disk has been replaced sometimes the resilver process can take an eternity. We have also found the snapshot replication process can interfere with the resilver process - resilver gets stuck at 99% and never ends - so we end up stopping or only doing one replication a day until the resilver process is done. The last helpful hint I have was lowering all the drive timeouts, see http://everycity.co.uk/alasdair/2011/05/adjusting-drive-timeouts-with-mdb-o= n-solaris-or-openindiana/ for info. thanks, liam On Sun, Mar 6, 2016 at 10:18 PM, Fred Liu <fred.fliu@gmail.com> wrote: > > > 2016-03-07 14:04 GMT+08:00 Richard Elling <richard.elling@gmail.com>: > >> >> On Mar 6, 2016, at 9:06 PM, Fred Liu <fred.fliu@gmail.com> wrote: >> >> >> >> 2016-03-06 22:49 GMT+08:00 Richard Elling < >> richard.elling@richardelling.com>: >> >>> >>> On Mar 3, 2016, at 8:35 PM, Fred Liu <Fred_Liu@issi.com> wrote: >>> >>> Hi, >>> >>> Today when I was reading Jeff's new nuclear weapon -- DSSD D5's CUBIC >>> RAID introduction, >>> the interesting survey -- the zpool with most disks you have ever built >>> popped in my brain. >>> >>> >>> We test to 2,000 drives. Beyond 2,000 there are some scalability issues >>> that impact failover times. >>> We=E2=80=99ve identified these and know what to fix, but need a real cu= stomer at >>> this scale to bump it to >>> the top of the priority queue. >>> >>> [Fred]: Wow! 2000 drives almost need 4~5 whole racks! >> >>> >>> For zfs doesn't support nested vdev, the maximum fault tolerance should >>> be three(from raidz3). >>> >>> >>> Pedantically, it is N, because you can have N-way mirroring. >>> >> >> [Fred]: Yeah. That is just pedantic. N-way mirroring of every disk works >> in theory and rarely happens in reality. >> >>> >>> It is stranded if you want to build a very huge pool. >>> >>> >>> Scaling redundancy by increasing parity improves data loss protection b= y >>> about 3 orders of >>> magnitude. Adding capacity by striping reduces data loss protection by >>> 1/N. This is why there is >>> not much need to go beyond raidz3. However, if you do want to go there, >>> adding raidz4+ is >>> relatively easy. >>> >> >> [Fred]: I assume you used stripped raidz3 vedvs in your storage mesh of >> 2000 drives. If that is true, the possibility of 4/2000 will be not so l= ow. >> Plus, reslivering takes longer time if single disk has bigger >> capacity. And further, the cost of over-provisioning spare disks vs raid= z4+ >> will be an deserved >> trade-off when the storage mesh at the scale of 2000 drives. >> >> >> Please don't assume, you'll just hurt yourself :-) >> For example, do not assume the only option is striping across raidz3 >> vdevs. Clearly, there are many >> different options. >> > > [Fred]: Yeah. Assumptions always go far way from facts! ;-) Is designing > a storage mesh with 2000 drives biz secret? Or it is just too complicate = to > elaborate? > Never mind. ;-) > > Thanks. > > Fred > > >> >> > *illumos-zfs* | Archives > <https://www.listbox.com/member/archive/182191/=3Dnow>; > <https://www.listbox.com/member/archive/rss/182191/25482196-63d208bc>; | > Modify > <https://www.listbox.com/member/?member_id=3D25482196&id_secret=3D2548219= 6-28027d72> > Your Subscription <http://www.listbox.com>; >
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?CAESZ%2B_-%2B1jKQC880bew-maDyZ_xnMmB7QxPHyKAc_3P44%2Bm%2BuQ>