BRRABill's Field Report With XenServer
-
@scottalanmiller said:
@coliver said:
For the most part a agree with this. Windows Admin already have or are aware of the tools to manage Hyper-V where as XenServer and ESXi require a new set of tools.
We were discussing this offline or in another thread. I'm not so convinced that this is a broadly true as people think. Tons of Windows Admins are unfamiliar with the tools you would expect them to know for this and even ones that use those tools, often they don't use them for Hyper-V and just log in through RDP to manage it. They might be familiar with the tools (maybe) but often don't even leverage them.
So true in my experience.
Learned a lot in the last couple years through VM's and the various and sundry things you need to poke to get them to work / manage them.
-
@FATeknollogee said:
@FATeknollogee said:
Not to side track this thread (apologies to @BRRABill ), what is the "hyperconverged" equivalent in the XenServer world?
To all you XS experts, what is the "hyperconverged" equivalent in the XenServer world?
Similar to Starwind in the Windows world
XenServer is natively that in the Xen world. Nothing additional needed.
-
@coliver said:
@FATeknollogee said:
@FATeknollogee said:
Not to side track this thread (apologies to @BRRABill ), what is the "hyperconverged" equivalent in the XenServer world?
To all you XS experts, what is the "hyperconverged" equivalent in the XenServer world?
Is Scale an example of hyperconverged? I don't think there is a single vendor that does that with XenServer.
Yes, Scale is one of the big HC vendors. They use KVM for their solution rather than Xen. KVM is popular for integrating into products, Xen is not. Xen is more popular in clouds than KVM. Partially just culture. Partially because KVM is much more capable of being altered for things like custom storage stacks.
-
@coliver said:
Hyperconverged just seems like a marketing term to me.
In many ways it is, but it does mean something.
-
@BRRABill said:
I would like to run a test of restoring my failing server to XenServer, but doing it without putting it on the network.
This possible with Hyper-V. You simply remove the network connection, and the can ... RDP I guess? ... on the local server into the VM.
Is this possible with XenServer?
It's exactly the same, just restore it and look at the console.
-
@BRRABill said:
Though I wonder if setting up a new VM and doing a bare metal restore of that server wouldn't be better than importing the VHD.
Still, pretty awesome either way.
One of the hurdles this would get me over is the too small size of my system disk. The VHD files really don't help since these are Server 2003 boxes. If can be as big as it wants, but the server still sees the smaller size.
I know I could use GParted or something like that, too.
-
@BRRABill said:
@DustinB3403 said:
Yes you can import the VM into Xenserver and remove the NIC from it.
You'd then administer the VM from within XenCenter or Xen Orchestra, it's simply disconnected from the network.
Just like if the cord was unplugged.
But can I actually SEE it? Like get the GUI of the server up on my screen?
yes, just look at XenCenter, the interface is so close to VMware's and HyperV's that I can't tell them apart at quick glance. There is a console display right in there.
-
@BRRABill said:
Remember I'm from the old RDP way of managing.
RDP doesn't work in this situation no matter what solution you are using. No different on Hyper-V.
-
@BRRABill said:
@Dashrender said:
I followed this to create my ISO Library locally on the XS datastore
http://www.riverlite.co.uk/blog/xenserver-creating-a-local-iso-library/(shivers)
Just share a folder from your desktop that you are running XenCenter on. Same as sharing files anywhere in the Windows world. Super simple, all Windows standard tools.
-
@BRRABill said:
@BRRABill said:
@Dashrender said:
will XenServer boot Windows based VHDs?
Yes, it did after I used the "fixup" option.
Though I wonder if setting up a new VM and doing a bare metal restore of that server wouldn't be better than importing the VHD.
Still, pretty awesome either way.
Yes, that is the proper way to do it both from a XenServer and from a StorageCraft perspective.
-
@BRRABill said:
@BRRABill said:
Though I wonder if setting up a new VM and doing a bare metal restore of that server wouldn't be better than importing the VHD.
Still, pretty awesome either way.
One of the hurdles this would get me over is the too small size of my system disk. The VHD files really don't help since these are Server 2003 boxes. If can be as big as it wants, but the server still sees the smaller size.
I know I could use GParted or something like that, too.
Bare metal restoring is what I did for my server. Took some effort to get things working.. but it did work.
-
@scottalanmiller said:
@FATeknollogee said:
@FATeknollogee said:
Not to side track this thread (apologies to @BRRABill ), what is the "hyperconverged" equivalent in the XenServer world?
To all you XS experts, what is the "hyperconverged" equivalent in the XenServer world?
Similar to Starwind in the Windows world
XenServer is natively that in the Xen world. Nothing additional needed.
If you had 2, 3 or more XS bare metal installs with local drives, how do you "hyperconverge" all the local disks?
Are you saying with XS the "hyperconvergence" just auto-magically happens?
-
@FATeknollogee said:
@scottalanmiller said:
@FATeknollogee said:
@FATeknollogee said:
Not to side track this thread (apologies to @BRRABill ), what is the "hyperconverged" equivalent in the XenServer world?
To all you XS experts, what is the "hyperconverged" equivalent in the XenServer world?
Similar to Starwind in the Windows world
XenServer is natively that in the Xen world. Nothing additional needed.
If you had 2, 3 or more XS bare metal installs with local drives, how do you "hyperconverge" all the local disks?
Are you saying with XS the "hyperconvergence" just auto-magically happens?
Of course not, but it doesn't for any platform. If you're setting up a greenfield situation, then you design it from the ground up with XS with single shared storage.
-
@scottalanmiller said:
@coliver said:
Hyperconverged just seems like a marketing term to me.
In many ways it is, but it does mean something.
So you say it means something but don't tell us what that is.
Please share.
Also, when should you and when shouldn't you care?
-
I'll go out on a limb and try answering my own question.
@Dashrender said:
Also, when should you and when shouldn't you care?
It's great to start with something like Scale when you're greenfield or when you're standing up a lot of additional resources, but after the fact - why bother? If you have new goals you need to achieve with the old platform, sure, spend the resources designing and moving to a new hyperconverged setup.
But if you don't have new goals, why would/should you spend the money/time changing? -
Now I'm ready for Scott to tear my assumptions to shreds.
-
@FATeknollogee said:
If you had 2, 3 or more XS bare metal installs with local drives, how do you "hyperconverge" all the local disks?
Are you saying with XS the "hyperconvergence" just auto-magically happens?
Not magic, no. But the needed tools are all there. For two hosts (and in some cases a few more) you use DRBD technology. That's the small scale stuff.
Once you want to go bigger you move to either Gluster or CEPH, both built in. If you want another tool you can do it, but you have those two enterprise options right out of the gate and they are at the forefront of technologies being used in massive scale systems of this nature, both mature and both very heavily tested.
-
@Dashrender said:
@FATeknollogee said:
@scottalanmiller said:
@FATeknollogee said:
@FATeknollogee said:
Not to side track this thread (apologies to @BRRABill ), what is the "hyperconverged" equivalent in the XenServer world?
To all you XS experts, what is the "hyperconverged" equivalent in the XenServer world?
Similar to Starwind in the Windows world
XenServer is natively that in the Xen world. Nothing additional needed.
If you had 2, 3 or more XS bare metal installs with local drives, how do you "hyperconverge" all the local disks?
Are you saying with XS the "hyperconvergence" just auto-magically happens?
Of course not, but it doesn't for any platform. If you're setting up a greenfield situation, then you design it from the ground up with XS with single shared storage.
Let's try this again:
In Windows, you can take multiple boxes, add Starwind or Datacore = hyperconverged using local storage (no SAN needed).
How do you do the same thing with XS?
-
@Dashrender said:
@scottalanmiller said:
@coliver said:
Hyperconverged just seems like a marketing term to me.
In many ways it is, but it does mean something.
So you say it means something but don't tell us what that is.
Please share.
Also, when should you and when shouldn't you care?
I'm just letting you sit and wonder.
-
@Dashrender said:
I'll go out on a limb and try answering my own question.
@Dashrender said:
Also, when should you and when shouldn't you care?
It's great to start with something like Scale when you're greenfield or when you're standing up a lot of additional resources, but after the fact - why bother? If you have new goals you need to achieve with the old platform, sure, spend the resources designing and moving to a new hyperconverged setup.
But if you don't have new goals, why would/should you spend the money/time changing?Well you don't change for the sake of change, that's never the answer. You change when the change is more beneficial than the cost of changing, of course.
But why is it beneficial as a concept given apples to apples situations? The goal of HC is that the components are together and managed together as a single entity. This isn't always the right answer, of course, nothing is. But there is a huge amount of benefit to bringing the storage and the computer under the same umbrella with heavy coupling. Remember that everything we have always done has been hyperconverged... any normal server is HC. Every stand alone box with local disks is HC, just on a single node scale. Breaking out the storage is costly, complex and cumbersome. The same holds true as we scale up. What makes sense with one node mostly holds as making sense with more of them.
