New thread, the previous was about kernel-level clustering, not layered 
software.

> On Dec 23, 2014, at 2:15 PM, Miles Fidelman via illumos-discuss 
> <[email protected]> wrote:
> 
> It really is a pain.

HA is hard. DR is hard. HA + DR is hard.

> 
> In the Linux world, Xen or KVM + DRBD + Pacemaker + CRM gives you a 
> relatively painless HA cluster.

This is block-level replication.

> In the BSD world, there's HAST + CARP get you pretty far along the same path.

Also block-level replication.

> But there's nothing equivalent in the illumos world (Xen is pretty thin too).

The illumos equivalent for block-level replication is AVS. This works ok for 
small systems.
Nexenta even productized it under the moniker "Simple HA"  Unfortunately it 
doesn't work
well for systems in the TB+ range due to the architectural issues of moving 
large amounts
of data and the speed of light. Few illumos vendors work on small systems and 
even
Nexenta killed "Simple HA" 

Today, most vendors do something other than block-level replication for large 
scale HA & DR.
Most people use dataset replication or application-level redundancy in the 
illumos world.
There are commercially-supported, general-purpose clustering solutions, such as 
RSF-1.

If you want to roll your own, I'd suggest pacemaker/heartbeat and znapzend as 
good open-source
starting points.

> 
> Kind of a large gap when it comes to using illumos in any kind of production 
> environment.

I don't see a large gap in capabilities. Many thousands of illumos-based 
systems are in 
production with HA and DR.
 -- richard



-------------------------------------------
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

Reply via email to