On 7/23/14, 1:38 PM, Miles Fidelman via illumos-discuss wrote: > Saso Kiselkov wrote: >> On 7/23/14, 12:57 PM, Miles Fidelman via illumos-discuss wrote: >>> Hi Folks, >>> >>> Is there anything like DRBD or HAST for illumos? Seems like ZFS + DRBD >>> would be really sweet for high availability clusters - but so far, none >>> of the pieces are together in one place. >>> >>> Trying to get to a small cluster w/ Xen+ZFS+HA >>> >>> ZFS: illumos, FreeBSD, not really stable anywhere else >>> Replication: DRBD for Linux, HAST for BSD*, ??? for illumos >>> Xen: Linux, NetBSD >>> >>> Sigh... >> AFAIK there's no native remote block-level synchronous replication >> solution for Illumos. >> >> For async replication, incremental snapshots are the way to go. Just >> transfer an incremental snapshot every 30-60s. They're relatively >> inexpensive and robust, so pool corruption doesn't immediately get >> propagated to the replication pair. >> >> For sync replication I personally use SAS fabrics and mirrored pools >> (conceptually, this is essentially what sync replication does, only over >> a special protocol). Passive SAS cabling can run 3m (so rack-to-rack is >> possible) and 20m with active cabling (room-to-room even). If longer >> links are needed I'd opt for iSCSI. At Nexenta we've got some early >> stuff in ZFS that allows us to prioritize reads & writes to mirror >> sides, but it's still pretty new and not battle proven. >> >> I know these are conceptually not exactly the same as DRBD or HAST, but >> functionally it's identical. >> > > Thanks for the model. It's kind of overkill for what we're doing - tiny > (4-node) cluster, mostly for development, but we run mail and a list > manager in a HA-configuration. > > The basic stack, right now is: > > VM (Debian Dom1) > DRBD > LVM > software raid > Xen (Debian Linux Dom0) > > Pretty stable, except... if you're following the Linux world, and the > stuff w/ systemd (moving lots of core functions into a monolithic blob) > - instead of upgrading Linux, I've been seriously looking at migrating > to either a BSD or Solaris based stack. Complicated by having a couple > of older machines in our cluster that don't have hardware virtualization > extensions. > > Kind of what I'd LIKE to do is: > > VM (illumos if I can get all the mail, list, antivirus, etc. to run, > otherwise Linux) > <something like DRBD or HAST> > ZFS (exported ZVOL) > Xen (illumos-based Dom0) > > Can't seem to find a full set of comparable software in either the BSD* > or illumos worlds. Sigh...
For the VM stuff, I'd recommend having a look at SmartOS. Sure, you won't be able to do KVM on machines without HVM and EPT support, but you'll still be able to run zones on there and run your mail (latest postfix, postgrey, etc.), AV (latest clamav), web (latest apache, nginx, ruby, php, etc.), DB (latest mysql, postgres, etc.) and other stuff. All of these are available in packaged form from the Joyent repos, so no need to compile from source. And to insure disaster recovery, unless you need sync replication (and judging by your description of your environment as "tiny"), run an incremental zfs send/recv job and be done with it - if need be, you can even run it every 10s. -- Saso ------------------------------------------- illumos-discuss Archives: https://www.listbox.com/member/archive/182180/=now RSS Feed: https://www.listbox.com/member/archive/rss/182180/21175430-2e6923be Modify Your Subscription: https://www.listbox.com/member/?member_id=21175430&id_secret=21175430-6a77cda4 Powered by Listbox: http://www.listbox.com
