StarWind Software

Hi,
When we install StarWind VSAN, do we have a choice to mirror some machines and not others or everything that is added to the StarWind shared disks will be mirrored?
I have some servers that I want to copy on both servers, but some others that can't be mirrored.

Should I partition my disks and only configure some of those partitions in the StarWind Manager?

A little more informations on our systems:
We have 2 x Dell R730xd servers configured identically (1 x 14cores cpu/ 128gb ram/ 6TB SAS 10k/ 3.2TB SSD with H730 controller)

I'm going to install Windows Server 2016 Datacenter and add the Hyper V role.
I was planning on creating 2 disks, 1 x 6TB on the SAS disks and 1 x 3.2TB on the SSD disks.

1st VM: File Server (mirrored) (2.5TB on SAS)
2nd and 3rd VM: SQL Server (individual, since the 2nd will be in a Fail-over Cluster Instance) (1.2TB on SSD)
4th to 7th VM: RDS Server (individual, since I need 4 up at the same time) (120GB/ea on SAS)
8th: DC (individual) (120GB on SAS)
9th: Print (mirrored) (120GB on SSD)
10th: App server (mirrored) (120GB on SAS)
11th: Special RDS server converted from current VMWare (mirrored) (120GB on SAS)

Is this possible?
Is this bad practice?

When a VM server is mirrored, can I do maintenance on 1 and the mirror will take over meanwhile?
The mirrored VM can be added to a cluster or we have to create a new VM to have them clustered?

I've never installed StarWind VSAN, but I've watched a few videos showing how to do it and read a lot of your documentation.
Do you have a "lab environment" where I can experiment with the manager?
Thank you for your time,
Best Regards,

Hello Grypht,

When we install StarWind VSAN, do we have a choice to mirror some machines and not others or everything that is added to the StarWind shared disks will be mirrored?

StarWind soft is able to mirror just storage. All VMs in cluster should be located on virtual storage and they could migrate from host to host. It depends whether in VMware HA, vMotion features are enabled. In case you are going to run Hyper-V cluster it would also has a HA feature.

Should I partition my disks and only configure some of those partitions in the StarWind Manager?

You could create in StarWind Management Console a simple image device, put it on local storage, connect it via iSCSI Initiator and get a volume in disk management.

1st VM: File Server (mirrored) (2.5TB on SAS)
2nd and 3rd VM: SQL Server (individual, since the 2nd will be in a Fail-over Cluster Instance) (1.2TB on SSD)
4th to 7th VM: RDS Server (individual, since I need 4 up at the same time) (120GB/ea on SAS)
8th: DC (individual) (120GB on SAS)
9th: Print (mirrored) (120GB on SSD)
10th: App server (mirrored) (120GB on SAS)
11th: Special RDS server converted from current VMWare (mirrored) (120GB on SAS)

Is this possible?
Is this bad practice?

Yes, this is possible. In addition to that, keep your DCs out of cluster. My recommendation to have at least 2 DC machines on each host.

I've never installed StarWind VSAN, but I've watched a few videos showing how to do it and read a lot of your documentation.
Do you have a "lab environment" where I can experiment with the manager?

I recommend installing the trial version and play with StarWind on your environment:
https://www.starwindsoftware.com/customer-page

Thank you for those answers.
I watched several more videos during the week-end and I think I understand better how it works.

I do have a few more questions:
1- How the VM server and it's replicat works? (Can I put them directly in a Cluster or I have to create a VM Server on both hosts?)

2- How does it work If I have to do a maintenance? Can I switch the load to the replicat while I work on the VM server and than switch it back to the replicat and do the maintenance on it?

3- If I want to create a test environnement, can I create 2 VM servers in Hyper-V manager and install StarWind VSAN in them?

Thank you,

Hello Grypht,

1- How the VM server and it's replicat works? (Can I put them directly in a Cluster or I have to create a VM Server on both hosts?)

Cluster installation procedure:

1. Create a StarWind device on Host #1 in StarWind Management Console
2. Create a replica from Host #1 to Host #2
3. Connect devices in iSCSI Initiator
4. Inicialize the Volume in Disk Management and create a simple volume
5. Add disk in Cluster
6. Move existing VM on Cluster Shared Volume

Can I switch the load to the replicat while I work on the VM server and than switch it back to the replicat and do the maintenance on it?

You could run production from one host.

3- If I want to create a test environnement, can I create 2 VM servers in Hyper-V manager and install StarWind VSAN in them?

Yes, you can try to play and test environment on VMs with 2012 R2 and 3 available NICs (sync, iSCSI, Management)

Please do not hesitate to contact me whether you have any questions or need my assistance.

Hi,
I followed the document: https://www.starwindsoftware.com/starwi ... -v-cluster
Each host has 1 Sync NIC (1Gb crossover to 2nd host), 1 iSCSI NIC (1Gb crossover to 2nd host) and 1 "public" NIC (1Gb to switch).

I created the witness drive, 2 CSV drives and configured the replication in StarWind Management Console.
I setup the MPIO and the iSCSI Initiator and connected each drive in fail over mode with local path as active.
I initialized and formated the drives as GPT with NTFS file system.
I created a new cluster, configured the drives as CSV in the cluster manager, and created 2 VM in it:
Host 1, owner of CSV1, got Test VM1
Host 2, owner of CSV2, got Test VM2

If I shutdown normally a server, the other one takes all of the load without issues.
If I disconnect the power on a server to simulate a power event failure, the load on the failed server doesn't get transfered to the other server.
On the remaining server, I can see that the ownership of the CSV gets transfered to the right server, but the states becomes failed.
I have checked in the log file, but I don't understand what is causing the issue.

Since I'm using the "Free MCP Licence" How can I get help with this issue?
Thank you

*****************EDIT****************
I connected my 10gb NIC from host1 to host2 with a twinaxial cable and my sync speed seems to be limited at 1gb.
Is it because I kept my 1gb NIC for synchronisation?
In the task manager, I can see both sync NIC at 1gb transfer speed
*****************EDIT****************

******** New Edit *********
I removed the 1Gbps NIC from the synchronisation Interface and only left the 10 Gbps and I can see a transfer speed of 6.7Gbps
So I guess I can't mix 2 NICs with different speeds.
******** New Edit *********

My only issue left is if a server comes down hard, the other one can't take over the disk and vm properly, I have to bring the disk online and start the vm manually.
Thanks

Hello,

Could you confirm that Witness device is connected on both hosts in iSCSI Initiator via loop back (127.0.0.1) only?

Hi, I have checked a few times, but to make sure here are screenshots (sorry they are in french)

Host HyperV 5: Host HyperV 6: Does it seem OK?
Thank you

Hello
The configuration looks fine.
So, my recommendation is restart Failover Cluster service and watch the behavior of roles and storage during test.
If it help and all resourses will run properly — it might be DNS issue.

Also, please double check that you can ping cluster and partner node by their name and FQDN.

Hi,
What do you mean by " watch the behavior of roles and storage during test"?
I can ping the cluster IP and the 2 nodes IP from my workstation, the cluster nodes themselves and from a different server without losing any ping. I also tried both short names and FQDN names without losing a single packet.

I checked in the DNS and I can find the cluster and the 2nodes both in the direct and in the reverse zones on every DNS Servers we have.
Nslookup also works both on cluster and nodes names and IPs.

Not sure what else I can check DNS-wise.
Thank you for your help,

Hello Gtypht,

Could you restart Microsoft Cluster service when you power off one host and Failover does not work fine? Whether everything will start or no?

Grypht wrote:What do you mean by " watch the behavior of roles and storage during test"?

I meant to check does them operate properly.

Hi,
Thank you for the reply.

I got in touch with your support team and after removing everything and starting from scratch it started working perfectly.

I did 2 things differently (as far as I can tell)
- I had created a witness of 10gig but this time I created it with 1 gig (no L1 or L2 cache)
- In the cluster network, I configured the sync card not to allow cluster or network communication on that interface

Thank you for your time.

Grypht,

Thank you for questions.
Please do not hesitate to contact us whether you have any questions.