Discussions specific to the VMware vSphere hypervisor
Post Reply
lars@norstat.no
Expert
Posts: 109
Liked: 14 times
Joined: Nov 01, 2011 1:44 pm
Full Name: Lars Skjønberg
Contact:

Choosing destination host

Post by lars@norstat.no » Nov 04, 2013 9:03 am

Hi, had to change some things so this weekend i deleted all my replicas and replicated them again. I have choosen destination to be my disaster cluster with two ESXi servers in it ...

When i replicated the 73 machines, all where placed in the host DESX01 and no one in the host DESX01. Why is this ? As i can remember from the database in Veeam 6 you don't specify the cluster as the destination but the host directly. Probably because of som API limitation from Vmware. But why are all machines placed on a single host ? Do i need to have HA enabled or something ?

dellock6
Veeam Software
Posts: 5685
Liked: 1604 times
Joined: Jul 26, 2009 3:39 pm
Full Name: Luca Dell'Oca
Location: Varese, Italy
Contact:

Re: Choosing destination host

Post by dellock6 » Nov 04, 2013 12:06 pm

Single host selection was a limit of the past, now with Veeam you can specify the cluster. Anyway, probably the VM is simply registered on the first available host, HA and RDS will kick in when you will try to power on the replicated VMs. There is no need to balance the placement as long as VMs are powered down.

Luca.
Luca Dell'Oca
Principal EMEA Cloud Architect @ Veeam Software

@dellock6
https://www.virtualtothecore.com/
vExpert 2011 -> 2019
Veeam VMCE #1

Vitaliy S.
Product Manager
Posts: 22773
Liked: 1526 times
Joined: Mar 30, 2009 9:13 am
Full Name: Vitaliy Safarov
Contact:

Re: Choosing destination host

Post by Vitaliy S. » Nov 04, 2013 12:09 pm

....or you can manually move the VMs to the second ESXi host, jobs should continue running normally.

lars@norstat.no
Expert
Posts: 109
Liked: 14 times
Joined: Nov 01, 2011 1:44 pm
Full Name: Lars Skjønberg
Contact:

Re: Choosing destination host

Post by lars@norstat.no » Nov 04, 2013 12:14 pm

Dellock6, there is a need to balance the placement as long as (maybe) VMware has a "7 task at the same time" limit on NBD traffic to one host. But recent findings indicate otherwise.

Vitaliy, That's what i have been doing. But it would be better if some automatic placement took place. Even taking into account which jobs you are running frequently so that these replicas where places evenly on the target hosts as per the limit i mention above.

dellock6
Veeam Software
Posts: 5685
Liked: 1604 times
Joined: Jul 26, 2009 3:39 pm
Full Name: Luca Dell'Oca
Location: Varese, Italy
Contact:

Re: Choosing destination host

Post by dellock6 » Nov 04, 2013 12:19 pm

With 73 VM honestly I would use hotadd mode for replica more than NBD, by simply deploying some Veeam virtual proxy on destination clusters.
I'll leave to Veeam guys instead reply about the balance of the VMs, I don't know how their algorythm works.

Luca.
Luca Dell'Oca
Principal EMEA Cloud Architect @ Veeam Software

@dellock6
https://www.virtualtothecore.com/
vExpert 2011 -> 2019
Veeam VMCE #1

lars@norstat.no
Expert
Posts: 109
Liked: 14 times
Joined: Nov 01, 2011 1:44 pm
Full Name: Lars Skjønberg
Contact:

Re: Choosing destination host

Post by lars@norstat.no » Nov 04, 2013 12:29 pm

See some of the other treads i have posted in recently. a lot of people are reporting that Hotadd is essentially broken and generating massive IOPS load on the target SAN. I i hugly reduced replication time after starting using SAN/NBD and using much less resources doing so ...

Exchange have reduced replication time from about 15 minutes to 5 minutes ... That means 3 times more replicas a day ....

Gostev
SVP, Product Management
Posts: 24460
Liked: 3413 times
Joined: Jan 01, 2006 1:01 am
Location: Baar, Switzerland
Contact:

Re: Choosing destination host

Post by Gostev » Nov 04, 2013 8:20 pm

This reduction is mostly coming from the fact that hot add takes significant time to initialize and de-initialize (hot add and hot remove takes time), whereas NBD is effectively instant. So yes, for near-CDP replication with little changes from cycle to cycle, NBD is certainly a better choice.

As far as "Hotadd is essentially broken", this blanket statement does not really work, as the IOPS issue you are talking about does not seem to affect every environment, for example we are yet to reproduce it internally... thanks again for your tip in that other thread for something else for us to investigate.

lars@norstat.no
Expert
Posts: 109
Liked: 14 times
Joined: Nov 01, 2011 1:44 pm
Full Name: Lars Skjønberg
Contact:

Re: Choosing destination host

Post by lars@norstat.no » Nov 04, 2013 8:47 pm

Yes i believe you are right concerning hot add and near CDP replication.

And thank you for correcting me concerning my belief that i am alone in the world, a road far to easy to travel sometimes. :-)

But there is at least 100% reproducibility in my environment and maybe the post as you say can point you in the right direction. I have had problems with hot add before, but the underlying storage have always been the same.

Post Reply

Who is online

Users browsing this forum: Baidu [Spider], Google [Bot] and 30 guests