Caution: SqlServer Cluster Update patch from WSUS

Microsoft launched KB3194721 (for SqlServer 2012) and KB3194718 (for SqlServer 2014) this month though WSUS channel. If you choose to deploy them though WSUS, just as I did, you may face a bug on a cluster based SqlServer.

 

On the server with SqlServer installed as cluster, these patches just act as cancelled or failed. For digging the problem, I download the package from Microsoft Download Site and try to install them manually. The first step is decompressing. Files are placed on the largest drive (I dunno it based on the total or free space). In my case, one disk drive for a database role is chosen. Then the upgrading procedure starts, with a bug that is:

If the owner of this instance role of cluster is the same computer, the role will be automatically stopped during update process, and will be started later. But while the role is set to stop, the disk drive is offline with it. So, if that drive is the largest, sadly, the update procedure fails — source files of update package are missing.

 

Even you run these packages manually, not from WSUS, it’s still not possible to choose the folder to decompress. If some drive of database role of cluster is larger than all server self-owned disks, you may want to STOP THE NODE in cluster as a workaround.

When you fall in love with Spanned Volumes

In our company, our major productive servers are hosted in vSphere as Windows-based virtual machines. Some of them need very big disks. Through the largest disk space supported in vSphere is 16TB (or maybe more, i dunno), I’m not interested to create such a big file in VMFS. It will be really a pain your *** when you have to move such a big file among datastores, even among data facilities by using portable hard drives.

Creating smaller disks instead of a big one for every Windows-based VM, IMHO.

Fortunately, Windows support to merge small disks as a big one, using dynamic disk function. There are 4 kinds of volume can be supported in such a condition: Spanned, Striped, Mirrored and RAID-5. You can consider them like JBOD, RAID-0, RAID-1 and RAID-5 functions in a simple legacy storage device.

Because data protection is existed in the SAN, Mirrored and RAID-5 volumes are not considered at all. Striped will make it a little faster when lots of virtual disks fight for limited IOPS resources within one datastore. But there is no way to extend any stripped volume (https://technet.microsoft.com/en-us/library/cc770938.aspx).

So, the useless volume format in a traditional world, spanned volume, is the best solution, maybe the only one. Next time when you need to create big virtual disks for Windows in any virtualization environment, don’t forget to consider using Spanned Volumes.

[Obsoleted] Windows 10.0.14393 cannot work with WSUS, yet

After several weeks testing, I’m sad to tell you all that at least currently, Windows 10.0.14394 (aka 1607) still cannot work with WSUS.

Problem: If there is Windows related package push from WSUS, svchost processes on Windows clients keep crashing.

 

Check this for details: https://social.technet.microsoft.com/Forums/windows/en-US/31718a20-64dd-43f7-87d7-c86f03f74a4d/14393-windows-update-crash-every-minute?forum=win10itprogeneral

Add Windows 10 (1607/14393) boot files to WDS on Windows Server 2012 R2

I’m testing to add the boot.wim files from Windows 10 1607 (10.0.14393) installation discs to Windows Deployment Services on Windows Server 2012 R2.

Test 1: Add x86 and x64 boot files to WDS
Passed.

Test 2: Add drivers to boot files.
Failed with WDS client on Windows Server 2012 R2.
Succeeded with WDS client on Windows Server 2016 TP5 connected to the target WDS.

Test 3: Create discover images.
Passed but failed in Test 6.

Test 4: Create x64 capture image.
Passed.

Test 5: Boot from boot files.
Passed.

Test 6: Boot from discover images embedded in ISO files.
Failed with error: WdsClient: There was a problem initializing WDS Mode.
Try to use other images created from WDS clients of Windows Server 2016 TP5: failed in the same way.

Test 7: Boot from x64 capture image.
Passed.

Test 8: Capture an instance of Windows 10.0.14393.
Instance for test: Windows 10.0.14393.10 x86 with up-to-date Office 2016.
Passed.

Test 9: Deploy an instance of Windows 10.0.14393 through WDS.
Passed.

Div dynamic loading is not supported in Windows IoT (yet?)

Hi. I built an ASP.net side powered by SignalR recently. The function is easy that dynamically load from a URL provided from SignalR into a div from browser side. It works like a charm in IE, Edge and Chrome. The JavaScript written in the web page is simple:

  • SignalR 2.2.0 and jQuery 2.2.3 are selected;
  • Connect SignalR at beginning;
  • When SignalR disconnected, reconnect;
  • When SignalR event raised, call one method to pull the data from URL specified and set as content of a div; and,
  • Call the method mentioned above at beginning to load the initialization data.

Then, I need to do the same thing in one Raspberry Pi 3. I choose Windows IoT with UWP to build the basis. UWP app is quite easy, only one WebView control is embedded with the starting URL set to the page of that ASP.net website. According to only Insider Preview version can support Raspberry Pi 3, build 14295 is selected, which is the latest one can be found from official site.

After this app deployed and started in Raspberry Pi 3, it acts quite different. The initialization data is shown without any problem. But no more SignalR event is proceeded. By debugging on the server side, SignalR OnConnected, OnDisconnected and OnReconnected are called repeatedly and frequently.

Due to no more information can be grabbed from app running in Raspberry Pi 3, I have no clue at first. The good thing is I made a similar solution before. The only difference between these 2 apps, is the previous one use iframe instead of div. Despite of no reason observed, after I change the div into iframe and make the data surrounded by html and body tags which filling div previously , the problem vanishes away.

Office 2013/2016 Activation Issue with Windows 7 deployed by captured images

To make deployment more convenient, we use Windows Deployment Services in our company. All images related is not the originals. Usually, we install Office products into a clean Windows instance before capturing. That’s is really a good way to save lots of time in every deployment.

The problem found is: if Office 2013/2016 is installed with Windows 7, even after sysprep run, the captured image is still not working well. If this image is deployed on another hardware, not the same one which is used to create this image, Office cannot be activated by KMS located in our company and error is 0x4004F00D.

To avoid that: At the very end before capturing, sysprep audit mode for example, run OSPPREARM.exe from Office folder to remove all data related to activation. This file can be found in “Program Files\Microsoft Office\OfficeXX”. Of course, if Office 32bit is loved in 64bit OS, “Program Files (x86)” should be a good replacement.
After OSPPREARM run, you may want to run “cscript ospp.vbs /dcmid” from the same folder to check the result. “Not Found” displayed is predicted.

BTW, if Windows 10 is used instead, this step is not necessary at all.

UWP Developing: Don’t forget the firewall on IoT devices

After I deploy my UDP based application to Raspberry Pi 2 running Windows IoT 10.0.10586, this application receives nothing, just like running in a normal Windows computer.

I guess that maybe something related to firewall, but:
1 There is no article related to target that. As the time I’m writing this post, googling “Windows IoT firewall” returns nothing related.
2 There is no page for firewall configuration in Windows IoT utilities site.

Fortunately, a test proved that there is a built in Windows IoT.
By using PowerShell connection, I send the netsh command to create an exception of firewall. It accepted and the my application started to receive package after then.

Related command for reference:
netsh advfirewall firewall add rule name=”A name for this exception” dir=in action=allow protocol=UDP localport=12345

Hope this helps.

UWP Developing: Local network protocal stack issue

I’m working on developing a Universal Windows App (Windows 10 IoT) recently. My app includes a UDP server for receiving packages.

The issue I found is:
When the UWP running on local computer, the datagram sent by local console application cannot be received.

Normally, while developing a socket based network application, another application for making and sending test packages should be prepared. The problem is, for testing UWP application, either the tester should be prepared as UWP application, or this tester can only be used on another computer laying on the same network.