David Finster via smartos-discuss wrote:
SmartOS, by virtue of using local storage, doesn't have the same
options for HA at the VM level.
In the event of hardware failure, all bets are off as to what you can
get out of the failed host, so you'll have to rely on backups anyway
for a restore. In other solutions this would be mitigated by the use
of shared storage, but that option isn't recommended in SmartOS.
The way we provide HA across the business is through application level
HA. In the case of MSSQL, we run AlwaysOn availability groups
(clustering) and for our web servers we run multiple instances across
different hosts with a load balancer in front (or round robin DNS). By
doing this we can survive a host going down due to hardware failure.
For anything we can't put into HA this way, we have a script that
takes hourly ZFS snapshots and ships them either off to another host
or the backup server (depending on restoration priority).
This is definitely the way to go.
We do something very similar: if the application supports application
level HA, use it. We also have a lax rule that says if it's to run in a
zone, use application level HA.
For applications that don't support, or the business won't pay for (read
windoze stuff!), application level HA we have a SLA stating how much
data loss is acceptable and base ZFS replication intervals for the VMs
on that. Some replicate once a minute!
I can see why those using software encumbered with draconian licensees
like the VMware way of doing things, we were quoted something like $60K
in extra licenses to support application level HA for one MS-SQL based
application!
--
Ian.
-------------------------------------------
smartos-discuss
Archives: https://www.listbox.com/member/archive/184463/=now
RSS Feed: https://www.listbox.com/member/archive/rss/184463/25769125-55cfbc00
Modify Your Subscription:
https://www.listbox.com/member/?member_id=25769125&id_secret=25769125-7688e9fb
Powered by Listbox: http://www.listbox.com
