Fixes per PR review

Author: alkohli

articles/databox-online/azure-stack-edge-gpu-2110-release-notes.md

@@ -63,8 +63,8 @@ The following table provides a summary of known issues carried over from the pre
 
 | No. | Feature | Issue | Workaround/comments |
 | --- | --- | --- | --- |
-| **1.** |Azure Stack Edge Pro + Azure SQL | Creating SQL database requires Administrator access.   |Do the following steps instead of Steps 1-2 in [Create-the-sql-database](../iot-edge/tutorial-store-data-sql-server.md#create-the-sql-database). <ul><li>In the local UI of your device, enable compute interface. Select **Compute > Port # > Enable for compute > Apply.**</li><li>Download `sqlcmd` on your client machine from [SQL command utility](/sql/tools/sqlcmd-utility). </li><li>Connect to your compute interface IP address (the port that was enabled), adding a ",1401" to the end of the address.</li><li>Final command will look like this: sqlcmd -S {Interface IP},1401 -U SA -P "Strong!Passw0rd".</li>After this, steps 3-4 from the current documentation should be identical. </li></ul> |
-| **2.** |Refresh| Incremental changes to blobs restored via **Refresh** are NOT supported |For Blob endpoints, partial updates of blobs after a Refresh, may result in the updates not getting uploaded to the cloud. For example, sequence of actions such as:<ul><li>Create blob in cloud. Or delete a previously uploaded blob from the device.</li><li>Refresh blob from the cloud into the appliance using the refresh functionality.</li><li>Update only a portion of the blob using Azure SDK REST APIs.</li></ul>These actions can result in the updated sections of the blob to not get updated in the cloud. <br>**Workaround**: Use tools such as robocopy, or regular file copy through Explorer or command line, to replace entire blobs.|
+| **1.** |Azure Stack Edge Pro + Azure SQL | Creating SQL database requires Administrator access.   |Do the following steps instead of Steps 1-2 in [Create-the-sql-database](../iot-edge/tutorial-store-data-sql-server.md#create-the-sql-database). <ol><li>In the local UI of your device, enable compute interface. Select **Compute > Port # > Enable for compute > Apply.**</li><li>Download `sqlcmd` on your client machine from [SQL command utility](/sql/tools/sqlcmd-utility). </li><li>Connect to your compute interface IP address (the port that was enabled), adding a ",1401" to the end of the address.</li><li>Final command will look like this: sqlcmd -S {Interface IP},1401 -U SA -P "Strong!Passw0rd".</li>After this, steps 3-4 from the current documentation should be identical. </li></ol> |
+| **2.** |Refresh| Incremental changes to blobs restored via **Refresh** are NOT supported |For Blob endpoints, partial updates of blobs after a Refresh, may result in the updates not getting uploaded to the cloud. For example, sequence of actions such as:<ol><li>Create blob in cloud. Or delete a previously uploaded blob from the device.</li><li>Refresh blob from the cloud into the appliance using the refresh functionality.</li><li>Update only a portion of the blob using Azure SDK REST APIs.</li></ol>These actions can result in the updated sections of the blob to not get updated in the cloud. <br>**Workaround**: Use tools such as robocopy, or regular file copy through Explorer or command line, to replace entire blobs.|
 |**3.**|Throttling|During throttling, if new writes to the device aren't allowed, writes by the NFS client fail with a "Permission Denied" error.| The error will show as below:<br>`hcsuser@ubuntu-vm:~/nfstest$ mkdir test`<br>mkdir: cannot create directory 'test': Permission denied​|
 |**4.**|Blob Storage ingestion|When using AzCopy version 10 for Blob storage ingestion, run AzCopy with the following argument: `Azcopy <other arguments> --cap-mbps 2000`| If these limits aren't provided for AzCopy, it could potentially send a large number of requests to the device, resulting in issues with the service.|
 |**5.**|Tiered storage accounts|The following apply when using tiered storage accounts:<ul><li> Only block blobs are supported. Page blobs are not supported.</li><li>There is no snapshot or copy API support.</li><li> Hadoop workload ingestion through `distcp` is not supported as it uses the copy operation heavily.</li></ul>||
@@ -84,7 +84,7 @@ The following table provides a summary of known issues carried over from the pre
 |**19.**|NFS |Applications that use NFS share mounts on your device to write data should use Exclusive write. That ensures the writes are written to the disk.| |
 |**20.**|Compute configuration |Compute configuration fails in network configurations where gateways or switches or routers respond to Address Resolution Protocol (ARP) requests for systems that do not exist on the network.| |
 |**21.**|Compute and Kubernetes |If Kubernetes is set up first on your device, it claims all the available GPUs. Hence, it is not possible to create Azure Resource Manager VMs using GPUs after setting up the Kubernetes. |If your device has 2 GPUs, then you can create 1 VM that uses the GPU and then configure Kubernetes. In this case, Kubernetes will use the remaining available 1 GPU. |
-|**22.**|Custom script VM extension |There is a known issue in the Windows VMs that were created in an earlier release and the device was updated to 2103. <br> If you add a custom script extension on these VMs, the Windows VM Guest Agent (Version 2.7.41491.901 only) gets stuck in the update causing the extension deployment to time out. | To work around this issue: <ul><li> Connect to the Windows VM using remote desktop protocol (RDP). </li><li> Make sure that the `waappagent.exe` is running on the machine: `Get-Process WaAppAgent`. </li><li> If the `waappagent.exe` is not running, restart the `rdagent` service: `Get-Service RdAgent` \| `Restart-Service`. Wait for 5 minutes.</li><li> While the `waappagent.exe` is running, kill the `WindowsAzureGuest.exe` process. </li><li>After you kill the process, the process starts running again with the newer version.</li><li>Verify that the Windows VM Guest Agent version is 2.7.41491.971 using this command: `Get-Process WindowsAzureGuestAgent` \| `fl ProductVersion`.</li><li>[Set up custom script extension on Windows VM](azure-stack-edge-gpu-deploy-virtual-machine-custom-script-extension.md). </li><ul> |
+|**22.**|Custom script VM extension |There is a known issue in the Windows VMs that were created in an earlier release and the device was updated to 2103. <br> If you add a custom script extension on these VMs, the Windows VM Guest Agent (Version 2.7.41491.901 only) gets stuck in the update causing the extension deployment to time out. | To work around this issue: <ol><li> Connect to the Windows VM using remote desktop protocol (RDP). </li><li> Make sure that the `waappagent.exe` is running on the machine: `Get-Process WaAppAgent`. </li><li> If the `waappagent.exe` is not running, restart the `rdagent` service: `Get-Service RdAgent` \| `Restart-Service`. Wait for 5 minutes.</li><li> While the `waappagent.exe` is running, kill the `WindowsAzureGuest.exe` process. </li><li>After you kill the process, the process starts running again with the newer version.</li><li>Verify that the Windows VM Guest Agent version is 2.7.41491.971 using this command: `Get-Process WindowsAzureGuestAgent` \| `fl ProductVersion`.</li><li>[Set up custom script extension on Windows VM](azure-stack-edge-gpu-deploy-virtual-machine-custom-script-extension.md). </li><ol> |
 |**23.**|GPU VMs |Prior to this release, GPU VM lifecycle was not managed in the update flow. Hence, when updating to 2103 release, GPU VMs are not stopped automatically during the update. You will need to manually stop the GPU VMs using a `stop-stayProvisioned` flag before you update your device. For more information, see [Suspend or shut down the VM](azure-stack-edge-gpu-deploy-virtual-machine-powershell.md#suspend-or-shut-down-the-vm).<br> All the GPU VMs that are kept running before the update, are started after the update. In these instances, the workloads running on the VMs aren't terminated gracefully. And the VMs could potentially end up in an undesirable state after the update. <br>All the GPU VMs that are stopped via the `stop-stayProvisioned` before the update, are automatically started after the update. <br>If you stop the GPU VMs via the Azure portal, you'll need to manually start the VM after the device update.| If running GPU VMs with Kubernetes, stop the GPU VMs right before the update. <br>When the GPU VMs are stopped, Kubernetes will take over the GPUs that were used originally by VMs. <br>The longer the GPU VMs are in stopped state, higher the chances that Kubernetes will take over the GPUs. |
 |**24.**|Multi-Process Service (MPS) |When the device software and the Kubernetes cluster are updated, the MPS setting is not retained for the workloads.   |[Re-enable MPS](azure-stack-edge-gpu-connect-powershell-interface.md#connect-to-the-powershell-interface) and redeploy the workloads that were using MPS. |