StarWind Software

We are currently running 5.7 and was waiting for 5.8 so that we can implement HA. What would be the optimal steps to upgrade to 5.8 and go HA with minimal downtime? We have created our existing image files with 65536 header size already.

Thanks,
Ara

Hello logicmate,

First of all I`d like you to know that the header is no longer lacated inside the *img file since version 5.8.
Could you please clarify if you are using HA or Basic Image target?

All images are currently setup as basic image target.

Ara

Well, then I think the best way will be like follows:

1.Destroy the target (keep the image file);
2.Create HA basing on the image file using 5.7 (please note that you need to perform synchronization in proper direction);
3.Update StarWind service on the first HA node. Wait until the service starts. At this step the HA node is unable to synchronize its HA devices and is not accepting the client connections. Client requests are processed by the second HA node. The next step will disconnect the existing client connections.
4.Update the StarWind service on the second HA node. Wait until the service starts.
5.Start synchronization on the first HA node. The second node changes its state to “ready” and starts processing client connections. Now, you can safely connect to the HA device.
6.Wait for synchronization to finish. Now the first HA node can process the client connections too.

Perfect, this is assuming that the new 2nd HA node has 5.7 already installed. Can we just start with 5.8 on the new 2nd HA node or we must create the HA using 5.7 on both HA nodes first?

Thanks,
Ara

You need to create HAs using 5.7.

So I installed 5.7 on the Partner node and did some tests (haven't upgraded to 5.8 yet) - Works fine so far.

1. I Destroyed target (kept the file HA-storage.img), Created HA target with current server virtual disk with HA-storage.img file and partner server virtual disk with HA-storage-partner.img "create new". When I click "next" it gives me the following message "The selected image is not HA compatible. All data on the image will be lost! Do you really want to use this image?". When I clicked yes everything was setup and completed. For this example I used ESXi for the host, I had to add the DataStore again (did not recognize the existing one), however it said something about data already existing and signature does not match, I selected the option "Keep the existing signature" and everything worked fine.

a)Is this the right procedure?
b)is the "selected image is not HA compatible" message okay or I have done something wrong?
c)is the warning I got from ESXi normal regarding the disk signature?

2. When the target alias for example is called HA-storage, can the partner target alias be HA-storage also? or it has to be named differently such as HA-storage-partner? Trying to understand if we should name them the same or different?

3. When creating the partner server virtual disk, is it best practice to name the file the same as the current server or should we append "-partner" at the end of the file like the examples? I know this shouldn't make a difference but just want to make sure we are using the best practice.

4. When we install 5.8, if HA-Node-1 has 2x1GB NICs for iSCSI client and HA-Node-2 has 2x1GB NICs for iSCSI client that would be 4x1GB NICSs for iSCSI clients total (theoretically 4000mbs total bandwidth with MPIO). How many NICs should be used for the sync channel? would we need 4x1GB NICs per server for sync? or 2x1GB per server is enough?

thanks,
ara

logicmate wrote:1. I Destroyed target (kept the file HA-storage.img), Created HA target with current server virtual disk with HA-storage.img file and partner server virtual disk with HA-storage-partner.img "create new". When I click "next" it gives me the following message "The selected image is not HA compatible. All data on the image will be lost! Do you really want to use this image?". When I clicked yes everything was setup and completed. For this example I used ESXi for the host, I had to add the DataStore again (did not recognize the existing one), however it said something about data already existing and signature does not match, I selected the option "Keep the existing signature" and everything worked fine.

a)Is this the right procedure?
b)is the "selected image is not HA compatible" message okay or I have done something wrong?
c)is the warning I got from ESXi normal regarding the disk signature?

Yes, you did everything correctly. The best thing that can proove it is that the data is OK, and as I get it is, correct?

logicmate wrote:2. When the target alias for example is called HA-storage, can the partner target alias be HA-storage also? or it has to be named differently such as HA-storage-partner? Trying to understand if we should name them the same or different?

The target name can be whatever you want, but the IQNs should be unique for every target (there shouldn`t be two identical IQNs in one system.

logicmate wrote:3. When creating the partner server virtual disk, is it best practice to name the file the same as the current server or should we append "-partner" at the end of the file like the examples? I know this shouldn't make a difference but just want to make sure we are using the best practice.

And you are correct again - it is really doesn`t matter how you`ll call it. Adding "-partner" part is useful because you shouldn`t remember where is partner and where is primary IMHO.

logicmate wrote:4. When we install 5.8, if HA-Node-1 has 2x1GB NICs for iSCSI client and HA-Node-2 has 2x1GB NICs for iSCSI client that would be 4x1GB NICSs for iSCSI clients total (theoretically 4000mbs total bandwidth with MPIO). How many NICs should be used for the sync channel? would we need 4x1GB NICs per server for sync? or 2x1GB per server is enough?

And correct again!:) Total bandwidth of Sync channel should be not less then total bandwidth of all the client connections.