StarWind Software

Is it possible to force a device to be identified as a SSD? It is very annoying that all my devices in vSphere is identified as a regular HDD. This is because the Megaraid-controller is not reporting as SSD to the OS Starwind is running on.

Shouldn't be very difficult, just a couple of NOP service commands added. I've fired this to RD as a feature request.

Thanks!
The main "pain" is that when having multiple ESXi-hosts accessing multiple LUNs/datastores, no hosts can use the LUN when marking as flash in the vSphere Web Client.
So I have migrated all VMs to other datastores, (formatted as VMFS6) and marked as flash on one and one hosts. This takes a long time. But be worst thing is that I have to migrate all VMs off again if adding a new host..
So this change will be great. If the storage is identified as SSD "all the way", we can make use of trim from the OS to shrink VMDKs and also LSFS-volumes.

It can be done in the current version of StarWind by adding "SolidStateDrive" parameter to device string in config file StarWind.cfg:

Start WordPad with administrator privileges, open StarWind.cfg file, find <device... record by value of device name or .swdsk file parameter.
Add SolidStateDrive="1" parameter before closing tag, for example:

<device file="My Computer\D\img\Storage4\Storage4.swdsk" name="imagefile1" SolidStateDrive="1"/>

Save file, restart StarWind service.

We will add this option to UI wizard in one of upcoming updates.

GREAT!
This can be done on HA-devices by changing on one node, sync and then the other node? To prevent downtime.

Let us check this in our lab. vSphere don't like to update information about "known" devices on fly. For example, it updates information about supported commands only after restart of ESX server.

OK.
Are you worried about a crash, or just that a reboot is needed? A reboot is no problem.

I tried this change on both nodes and rebooted host. No automatic change of disk type.. :\
Is this a change in R5, og should the current release from Aug support this setting?

I had doubts about need of ESXi reboot for refreshing device info.

It works with V8R4 from August 2016. We have checked with 3-node ESXi 6.5 cluster.
Option has been added to both StarWind HA nodes: edit StarWind.cfg, restart StarWind service, wait for HA sync, repeat for the second node.
Then all three EWXi nodes were restarted one-by one.
After storage resan ESXi showed disk type as Flash.

All this time test VM with test load remained functional.

Important remark: in case of HAImage device, we need to add the "SolidStateDrive" parameter to corresponding ImageFile device. This device is mentinoned in HAImage's "Storage" parameter, for example:
<device name="HAImage1" ... AluaNodeGroupStates="0,0" Storage="imagefile2"/>
For this device we need to locate "imagefile2" in config file to add the parameter:
<device file="My Computer\D\img\Storage5\Storage5.img" name="imagefile2" SolidStateDrive="1"/>


There is another method for marking storage as SSD on ESXi side:
https://kb.vmware.com/selfservice/searc ... Id=2013188

It might be that I have to reboot all hosts first. Will check.

btw, the ESXi-method sucks, as no host can use the datastore, not only the host I make the changes on. Not ideal when having 30-ish hosts.. :\

Sigh.. I did a reboot of all hosts, one and one, and still not identified as flash storage.

Did the change like this:

<device file="My Computer\F\p3600_1\p3600_1.swdsk" name="imagefile1" SolidStateDrive="1"/>

This should be the correct method?
Might be a delay in vCenter, but I don't think so.

Yes, that's right!
You can check StarWind log file for the string "disk is SSD" near the records about initialization of ImageFile device.

What ESXi version are you using?
Does datastore rescan or "refresh capacity information" change anything?

Will check logs. I have tried rescans "here and there" without any luck.
Running vCenter 6.5 and ESXi 6.5

From log:
G:\p3600_2\p3600_2.img': disk is SSD, UNMAP supported

So I will have to look at the ESXi side. Very strange.. ALL hosts have been restarted, but not simultaneously of course (NOT possible).