Virtualization Redemption?
-
Here are the heavier details on HyperV Replication from Technet. Pretty sure you can choose Azure as the DR too, if you wanted.
-
Also... more thinking out loud... Doesn't XenServer have replication options like that as well?
-
@dafyre said:
Also... more thinking out loud... Doesn't XenServer have replication options like that as well?
That's what DRBD Proxy is. They've had it long before HyperV did, but it is less than trivial to set up.
-
@scottalanmiller said:
@dafyre said:
Also... more thinking out loud... Doesn't XenServer have replication options like that as well?
That's what DRBD Proxy is. They've had it long before HyperV did, but it is less than trivial to set up.
I was thinking more along the lines of like Hyper-V's replication... not the DRBD / HA setup.
-
Okay, so it's possible we have reached a free (other than my time) means of accomplishing this goal. yes?
I can do all of this on site so tell me if i'm off base or not.
-
setup new DR server with 2 new drives and reconfigure to raid 6
-
Setup Hyperv12 on DR server
-
P2V current VMs to DR host temporarily
-
HV 2012 on two "main site" servers
-
Migrate servers to their respective hosts
-
Setup Replication locally before shipping server to datacenter
-
Ship DR server to datacenter
-
Allow replication to do its thing?
-
-
@hubtechagain That sounds about right. You should note that the replication is not real time and happens on a set schedule (not sure what the lower limit is), so there could be a potential for data loss of XX minutes between the time HOST1 dies and its last replication to HOST_DR.
I would also dare suggest that Replication is not a replacement for backups. 8-)
-
from TechNet
Replication frequency—In Windows Server 2012 replication occurs every 5 minutes. In Windows Server 2012 R2, you can configure the replication frequency every 30 seconds, 5 minutes, or 15 minutes.That being said, this server will only ever be used if we have a complete physical failure at our main location. this is EXACTLY what i was looking for @dafyre you get atleast 100 nuggets.
-
@hubtechagain said:
-
setup new DR server with 2 new drives and reconfigure to raid 6 (if that means 5+ drives per machine, then yes that makes sense.)
-
Setup Hyperv12 on DR server (Yes!)
-
P2V current VMs to DR host temporarily (Yes!)
-
HV 2012 on two "main site" servers (Yes!)
-
Migrate servers to their respective hosts (Yes!)
-
Setup Replication locally before shipping server to datacenter (probably makes sense.)
-
Ship DR server to datacenter (Yes!)
-
Allow replication to do its thing? (Yes!)
-
-
Sounds like a good plan. Now the question becomes, how do you handle the two local servers? Are they going to be stand alones? or are they going to do full clustering like StarWind?
-
@dafyre said:
@hubtechagain That sounds about right. You should note that the replication is not real time and happens on a set schedule (not sure what the lower limit is), so there could be a potential for data loss of XX minutes between the time HOST1 dies and its last replication to HOST_DR.
I would also dare suggest that Replication is not a replacement for backups. 8-)
DR sites, currently, are always "non real time" or asynchronous. No reasonable technology lets you do anything differently today. HyperV Replication, Veeam, DRBD Proxy... all the same.
-
@scottalanmiller They will more than likely just be two individual hosts. why would I change that up now?
-
@scottalanmiller If he could get it at 30 seconds, that would be acceptable! Especially in a full DR scenario, lol.
Although Scott does have a good point... How are you going to deal with the servers locally? (Local Backup, or Starwind?)
If you are on Spiceworks, you can get a 2-node unlimited storage Starwind license for free, I think.
-
considering we were willing to lose a day of data to our DR site.... this will replicate much more often, so this is already better than what i've been thinking.
-
WTF is starwind?! ha
-
lol. Starwind basically lets you take 2 servers and build a SAN out of them... Real time replication across the 2 hosts and all that fun stuff... So if you have 2TB of storage in each server, it builds you a SAN with 2TB of storage (think Networked RAID-1)...
However, if your servers are now standalone, and you are happy with that... It may not be worth rocking the boat...
Thoughts, @scottalanmiller ?
-
@hubtechagain said:
@scottalanmiller They will more than likely just be two individual hosts. why would I change that up now?
If you wanted HA. Now that you are leaving the ESXi world, the big features like vMotion and High Availability (at the hypervisor level) are free. It's a whole new world of opportunity.
-
@hubtechagain said:
considering we were willing to lose a day of data to our DR site.... this will replicate much more often, so this is already better than what i've been thinking.
But is it the best application of the resources at hand? Why do "good enough" when "better" is also an option?
-
@dafyre said:
@scottalanmiller If he could get it at 30 seconds, that would be acceptable! Especially in a full DR scenario, lol.
Thirty seconds instead of zero isn't big, but it is just one piece of the puzzle. Going to "users will never know" is nice.
Going from "a little data loss" to zero is bigger.
Going from "Crash consistent" to "fully consistent is bigger still."
-
@hubtechagain said:
WTF is starwind?! ha
Starwind is the leader in storage replication technology for both HyperV (all scales) and ESXi (below three nodes). They are free and the only major player in this particular space. @KOOLER
-
@dafyre said:
However, if your servers are now standalone, and you are happy with that... It may not be worth rocking the boat...
Thoughts, @scottalanmiller ?
Can go either way. Worth talking to the customer, though. This is a BIG move forward that they are getting for "free", other than Hub's time (which he presumably wants to sell more of) , by moving off of VMware. This is not just a chance for him to expand his value to the customers, but a chance for him to showcase a leap forward for them, rather than an incremental advance.
