While ‘resting between engagements’, I took the opportunity to install and configure a Solution Manager system on a cloud host local to Australia.  The main reason was for a demojam entry, but it’s always good to keep my skills up to date.  The target system provided was a Windows 2008 R2 system.

Windows
Server 2003 R2 and Windows Server 2003 only have functional differences; using the same SAP kernel version, the same service packs, and
the same hot fixes and security fixes.  By contrast, going from Windows Server 2008 to Windows Server 2008 R2 requires an updated kernel (see SAP note 1383873 – Windows Server 2008 R2 Support).  Now, I could say I was using 2008 R2 for all the right reasons; for example, according to Frequently Asked Questions – SAP on Windows Server 2008 R2:

The main benefits of Windows 2008 R2 are

• Scalability
  Windows Server 2008 R2 supports up to 256 logical processors.
• Improved virtualization features
  On Hyper-V in Windows Server
  2008 R2, the amount of cores supported by the hypervisor (up to 32) has
  been enhanced. Another enhancement is Live Migration support by the
  implementation of Cluster Shared Volumes (CSV). Virtual Machines can be
  migrated without service interruption between the cluster members.
• Power usage
  Windows Server 2008 R2 reduces processor power
  consumption in server computers with multi-core processors using a
  feature known as Core Parking. Core Parking allows Windows Server 2008
  R2 to consolidate processing onto the minimum number of required
  processor cores, and suspends inactive processor cores. The advantage of
  Core Parking over traditional servers is 10-15 % energy saving for the
  same workload.

For a complete list of features, see:
http://www.microsoft.com/windowsserver2008/en/us/overview.aspx

However the real reason was that Windows 2008 R2 was already installed on the server I was using.  This became a bit of an challenge, as the Install Kit I used wasn’t actually for 2008 R2 !!.  I found OSS Note 1383873 fairly quickly, but even after installing the appropriate kernel as suggested,  sapstartsrv.exe (used by the SAPxxx_NN service) would not start correctly.  I discovered via google that I needed to install an extra Microsoft c-runtime (vcredist) to run the new SAP kernel.

The reason for posting this as a blog (I’ll also add it to the wiki) is that while I’ve since found out that this is ‘general knowledge’, it wasn’t described in OSS Nnote 1383873 – Windows Server 2008 R2 Support , and in fact, the only reference I found to my symptom was in one line in 1494740- SAP system migration from Windows 2003 to 2008 (R2) 64-bit AFTER I chased the error down through Google.

Are you implementing or using Windows 2008 R2 ? If not, why not ? Corporate standards ?, Lack of product support ? Lack of in-house knowledge ?

An SAP Administrator needs to know about more than SAP; they need to know about the ecosystem that their systems run in.  By that, I mean things like how to use features of the Operating System and DBMS that their systems run on, to provide value to the system or business owner.  For example, you probably know that drivers, services or software in Windows can crash without you even being aware of it happening.  Sometimes this can affect your Solution Manager system, another non SAP part of your Landscape, or you may just want to monitor something like an NSP Developer Edition system.  Whichever type of system we are talking about, sometimes the first sign of trouble is when you (or even worse someone in the business) needs the system right now. What would be useful would be a tool that notifys you when certain activities occur……

The Windows Event Viewer lets you launch a program, send an email (if the server has an email client installed) or provide some other alert that something has occurred.  You do this by attaching a task to an Event in the Event Viewer.  To do this you need to find your Event within Event Viewer.  Note there are slight differences in the initial screens between Windows XP, Windows 7, Windows 2003 and Windows 2008.

Once you’ve found the Event you want to report on, look in the right hand panel.  There you will see an option Attach task to this Event.  Selecting this will pop up a window with all of our options.

For example, we can run a program, send an email (if email software is installed on the Server) or display a pop-up alert.

If you want to run a program, there are some very useful command Line and PowerShell utilities that can come in very handy here.  I won’t go into much detail as they are well documented on the Microsoft website, but examples include running the program CMD.EXE with either the /c switch to carry out a command string and then stop, or the /k switch to continue afterwards (see here for more details on command line switches).

You can also use the WEVTUTIL command to automatically poll the event viewer for data and perform actions like creating a log to the Administrator or <sid>adm desktop.  This would make it easier to send selected data to second level support or SAP.

You can also use PowerShell command to automatically generate a Windows System Health Report:

Get-RmsSystemHealthReport -Path <drive>:\Report [-StartTime <start_time>] [-EndTime <end_time>] -ReportType <report_type

Any tasks you add can be viewed and edited in the Windows Task Scheduler.  The important thing to remember is that being able to add actions to events can be a real time-saver when it comes to diagnosing problems in Windows.

What Operating System DBMS tools and scripts have you found useful for monitoring systems and software ? Are the Windows tools better thean the Unix / Linux tools ? What about agents for centralised monitors ? Which do you prefer ?

After my previous post, you either have your own Windows 64-bit AMI image, or access to the Public AMI I have created, called sap.nw70.win-64.db2. In this exercise, we will use this as the basis of a new, private, image that will:
* contain the appropriate installation data (including registery keys) for SAP NW7,
* be capable of online / offline backups, using SAP tools,
* provide a painless way of running 24×7.

Prerequisites

* EC2 and S3 Accounts with Amazon,
* access to a Solution Manager system (for the installation key),
* access to an OSS ID with download authorisation.

Architechture

Once we terminate an Amazon instance, we lose all changes to it. Saving our database and configuration changes by bundling the changed system into a new AMI will take a non trivial amount of time; Certainly enough to prevent it being run 24×7. Additionally we will lose lots of usefull ABAP and JAVA stack logs unless we bundle the running instance every time we shut it down.

Just as well there’s an alternative, called Elastic Block Storage. This allows you to create data volumes and mount them on your image. They are persistent, and more importantly, can be backed up by snapshots, from the AWS Management Console.

So that leads to an architechture (or rather, disk layout) as follows:

• Drive C: AMI instance, boot disk
• Drive D: AMI instance, ephemeral disk (data lost whenever instance shuts down)
• …
• Drive H: AMI instance, ephemeral disk (data lost whenever instance shuts down)
• Drive W: Persistent Disk, for storing disk-to disk backups DBMS and / or logs
• Drive X: Persistent Disk, for SAP and DB2 Intallation
• Drive Y: Persistent Disk, for DB2 logs
• Drive Z: Persistent Disk, for storing installation files

Creating EBS (Persistent) Volumes

To create EBS Volumes, go to the EBS Volumes section of the Amazon Management Console. The major issue with creating volumes is that you can only attach / mount an EBS volume on an instance that is running in the same Availability Zone. This does mean that all your volumes must be in the same Availability Zone, if they are to be attached to the same instance.

• Drive W: vol-a82bc7c1, for storing disk-to disk backups DBMS and / or logs
• Drive X: vol-3f658956, for SAP and DB2 Intallation
• Drive Y: vol-4451bc2d, for DB2 logs
• Drive Z: vol-fc2bcb95, for storing installation files

Note that these are empty, unformatted, unmounted, unattached volumes (at the moment…).

Attaching EBS Volumes to our Instance

To attach the volumes to an instance, we need to have an instance running. Start up an instance of your image or of sap.nw70.win-64.db2.

You assign a name to a volume when you are formatting it. You do this by running the Computer Management (if you’re running an instance of sap.nw70.win-64.db2, this should be on the Desktop of user sapinstall) and formatting and naming the volumes. Make the names distinctive, and related to their purpose, for example sw_repository.

Now use the C:\Program Files (x86)\Amazon\Ec2ConfigSetup\Ec2ConfigServiceSettings.exe program and the Drive Mapping tab to control which volume gets mounted to which drive letter. This is important, because we want to make sure that our sap_install, db2_logs, and backups volumes are always mounted on the same drives. Once the current image is bundled and registered, any instance launched from the new AMI will contain the setting we have configured in Ec2ConfigServiceSettings.exe.

System Specific Configuration

Change the hostname (or in Windows terms, the Computer Name) to one of your choosing (Start –> Control Panel — System –> Computer Name –> Change). Run Ec2ConfigServiceSettings.exe. and make sure the Set Computer Name flag and the Sysprep flag on the Syprep tab are disabled – They should already be disabled, if you are using a copy of sap.nw70.win-64.db2.

Check the swap space (Start –> Control Panel — System –> Advanced –> Performance Settings — Advanced, Virtual memory). Again, this should already be correctly set if you are using a copy of sap.nw70.win-64.db2.

Edit the hosts file in C:\windows\system32\drivers\etc to include your Computer Name as a valid host name, for internal SAP and DBMS connectivity.

Finally, bundle the instance using the AWS Management Console and register the resulting image under your own image name. The purpose here is to save the customisation you have done if you have a problem with the SAP installation. As part of the process of bundling, the instance is shut down and restarted.

For future reference, if you restart the instance yourself, using Start –> Shutdown and specifying Restart, you don’t loose any information or configuration from the C drive as you would if you terminated it from the AWS Management Console. This is because the later removes the underlying resources, while using Start –> Shutdown –> Restart doesn’t release the underlying resources.

Security and Firewalls

EC2 provides its own set of firewall rules called Security Groups. The defaults values are, essentially, just enough to get you access to the server itself.

Installation

Download the appropriate files from http://service.sap.com/swdc (you’ll need an S number with download authorisation), extract / expand them and store the results on the Z drive. I stored the download files under Z:\NW70SR3 and expaneded them into their own folders on the Z drive.

Make sure you read the appropriate OSS notes. For the ECC6 IDES, the important ones are:
0799639 – General IDES related
0956921 – NW7 ECC6 SR3 IDES related
1244548 – NW7 ECC6 SR3 IDES related
and
1126127 – DB6: Deferred Table Creation and Row Compression

Otherwise, the install follows the standard process, as detailed in the appropriate installation guide (in my case, the NW7.0 SR3 ABAP+JAVA / Windows/ DB2). The two exceptions are:
* Specify that the SAP and DBMS Installations go on an EBS volume (i.e drive X)
* in my case, specify that the DB2 logs go on an EBS volume (i.e. drive Y)

The full IDES install took around 30 hours run time (think of it as $20 or so well spent) from when I started sapinst (that time did include checking and amending my previous implementation notes). The majority of the time is spent loading about 150GB data into the DB2 database. However, once sapinst had accepted the Solution Manager Key, you can disconnect RDP and leave the install running.

Saving your image

Once the installation is complete, you’ll want to back it up before you go any further. Using the SAP MMC, shut down SAP (or logon to Windows as the SAPService<sid> user and shut down SAP).

Use the AWS Management Console to bundle your running instance.

Basically, the image consists of whats on the C Drive, so backing up your EBS Volumes requires you to use the AWS Management Console to save snapshots of them. The EBS volumes are stored and charged for at the Amazon S3 rates. Just like EC2, however, you are only charged fo what you use. This means that if you define a 500GB volume, write a 1 GB file to it and create 4 snapshots of the volume, you will charged for 5GB of storage; 1GB data on the volume, plus 4 lots of 1GB of snapshot. backup.

When you’re finished with the instance, shut down SAP and don’t forget to terminate tthe instance via the AWS Management Console (otherwise you’ll be charged for it !!).

Running your SAPSystem

Start an instance of your image and attach the EBS volumes to the running instance. The work of of assigning drive letters, in the correct order, to each volume is controlled by our configuration work earlier in Attaching EBS Volumes to our Instance. One of the issues currently outstanding is that thess will actually get mounted on subsequent restarts of this instance (which we perform below).

Logon to the instance and update / verify the Swap Space sttings via Start –> Control Panel — System –> Advanced –> Performance Settings — Advanced, Virtual memory.

Accessing your SAP System

Assuming you have opened the correct ports in the Security Group specified for this instance, you can now put the appropriate values into your SAP GUI …..

Whats next ?

You now have a running SAP system. However

• No DBA processing, i.e. no DB13 jobs, no backing up of logfiles etc has
  been implemented, so once you’ve tested connectivity, stop the SAP and
  DBMS systems and take snapshots of your SAP & Database volume.
• The SAP*, DDIC and IDADMIN passowrds are well known (or easily determined). Change them
• No post implementation work (i.e. SGEN) has been done,

The purpose of the exercise is to demonstrate how quickly you can run up a demonstration, training or testing system. Depending on how many resources you want to pay for (CPUs and memory), this can be quicker or slower.

However, it has been my experience, based on several green fields implementations, individual system implementations and upgrades, and feedback from others, that building an appropriate server – whether physical or virtual – can take up to 2 weeks. Using the approach detailed here, services such as provided by the Amazon EC2 service reduce this to the 45 minutes it takes to configure and bundle a standard public instance.

One of the obvious issues is that it is well and good using predefined data, which you can download, in zipped form, from OSS (such as the IDES data I used in this example). What about copying ‘real’ data fron an existing SAP system, especially if we’re talking TerraBytes ?

I’ll discuss this, the bandwidth of a portable hard disk and more of the Amazon Web Services features that are particularly useful for SAP in my next post.

UPDATE: I have tidied this up a bit, to make some things clearer and to include the name of an AWS Public Image that can be used as the source for the subsequent step.

In this post, I describe how I setup a windows environment to install SAP ABAP and Java stacks, using the Amazon Simple Storage Service (S3) to store persistent data.  I needed to:
* install and modify an appropriate Windows 2003 Server environment,
* save this environment for future use

In a subsequent post, I will describe the installation of an IDES system running NW7 and DB2.  The three major challenges were
* setting up persistent storage of the NW and DB2 installation,
* suitable for using standard SAP and AWS functionality to support sustained (i.e. 24×7) operation of the SAP system
* and allowing you to stop and start the SAP system and / or server without losss of persistent data.

The result is a fast and cheap way of running up multiple systems, with the following features:

* You are only charged running costs for those systems that are running
* Low running costs (at the time of writing, $US 50 cents an hour)
* Low storage costs ($US 15 cents / GB / month for your 50TB)
* No more waiting for hardware – you can start implementation right now
* Systems (i.e. extra application servers) can be implemented, but not running

What did I know I would need ?

After reading the NW 70 SR3 installation Guide for Windows / DB2, I knew the following:
* I needed a 64 bit Windows Server with authentication services,
* I needed a reasonable amount of RAM, plus a decent swap space,
* I needed JAVA 1.4.

After reading the AWS EC2 documentation, I also knew that it was not practical to keep any volatile datasets (i.e. DB2 itself, DB2 logs, SAP process logs, etc) as part of the server, and that I would need to use the Amazon EBS servcie for persistent storage.

Signing up for Amazon EC2 and S3

An excellent account of how to setup a Windows Server image, and the principles behind this, can be found at Dave Winer’s EC2 for Poets. It also gives a good overview of how to sign up for both EC2 and S3 and the issues around persistent data.

Creating the base Amazon Machine Image (AMI)

Logon to the AWS Management Console and select the Amazon EC2 tab.

Select the Launch Instance button…

.. then find and select the Basic 64-bit Microsoft Windows Server 2003 with Authentication Services image.

 Once the server shows up as running, logon using the techniques described in Dave Winer’s EC2 for Poets.  One of the first things I did was to create a sapinstall user.  This allows me to logon (via RDP) as user sapinstall / password without having to muck around with the keypairs. 

Changes to standard AWS Windows 2003 64-bit Image

There were five issues that needed to be dealt with.

First I had to disable the Windows Attachment Manager (for non-windows people, this is a security setting that Windows uses to stop you writing dangerous file types to your disk) before Internet Explorer would let me save files.  See the Microsoft Knowledge Base Article 883260 for a rundown on how it works.  The quickest way to disable it is to uninstall the Internet Explorer Enhanced Security Configuration. To do this, click Add or remove programs in Control Panel, click Add/Remove Windows Components, and then click to clear the Internet Explorer Enhanced Security Configuration check box.

2) Both SAP and DB/2 (my target DBMS) require that the host name of the server its installed and running on remains the same.  However, the default action every time you restart an AWS image is to have the host name set to IP-xxxxxx where xxxxxx represents the internal (to Amazon) host name the server is running on.

While you can perform arcane scripting to fix the host name, Amazon provide a tool, bundled within every AWS windows instance, that will ensure the hostname remains set to what ever you set in the System –&gt; properties screen.   The tool is C:\Program Files (x86)\Amazon\Ec2ConfigSetup\Ec2ConfigServiceSettings.exe

3) I wanted to make sure I had enough swap spacxe to run my SAP system.  The base instance we are using gives us 15GB of memory, but, especially if we want to install multiple JAVA engines, this may not be enough.  I allocated another 1500MB on each of two of the ephemeral disks.

4) My initial installation is going to be an NetWeaver 7 ECC6 system. This means we need to download and install java 1.4 from Sun’s Sekrit Squirrell place for old releases.  Don’t forget to setup the Environment variables (JAVA_HOME and PATH) correctly. 

5) The last change was to incorporate a Dynamic DNS Update tool.  This is used to pass the IP address of the server we are “running on” to a service that will then set a fixed Domain name to specify the same DNS name to users and tools whenever I ran my instance.  I use dyndns org.  You can register a limited number of domain names for free, and they provide a tool (DynDNS Updater) that allows you to register your IP address against one or more of your Domain names.

Save your Amazon Machine Image (AMI)

Now you have an instance you can use to install and run SAP on.  However, we need to make sure that all our changes are not lost.  This utdown means you need to “bundle” your running system into a standalone Amazon Machine Image.  Go to the Amazoin EC2 tab of the Amazon Management Console, select Instances, then select the instance you want bundled.  Right click on More Actions and select Bundle Windows AMI.

This generates a popup screen.  Fill out the appropriate details and clcik bundle. The Bundle Name refers to the S3 folder that will hold the AMI.  This must already exist.  The Key Name is appended to the name of manifest.xml filre that contains the S3 layout and location of your image.

 Once you click bundle your request is confirmed.

.

You can follow the progresss of the bundling by examining the Bundle Tasks screen.  There are three steps that bundling Windows instances needs to follow- The instance must shutdown, the Amazon bundling process must occur, and the resulting data must be stored.

Once the image has been bundled and stored, you must register the bundle as an Amazon machine Image.

An alternative to repeating all the work shown above is to grab a copy of the Public AMI I have created, called sap.nw70.win-64.db2.  You will need to change the hostname (as descibed above), implement your own DynDNS org domain name  and bundle and register the changed image.

Either way, you now have your own mildly customised image copy of a Windows 2003 Server, running on the  Amazon Web Services cloud.  This image is ready for installation of a non-trivial SAP system, such as the NW7 ECC6 IDES system.  

In the next post, I will describe how I used the sap.nw70.win-64.db2 image to install the Windows DB2 IDES for ECC6 system.

One of the sizing issues with an SAP system that doesn’t receive due consideration is the network capability; not just speed, but throughput. It’s always usefull to know what your Network is capable of, especially if you have lots of data to move (Support Packs / Support Stacks and so on). But how do we find out?

NetCPS (a single executable file) is rather simplistic, with no fancy features as the author (credits to Jarle Aase) says. It pumps 100MB of generated data (without accessing the HDD which could mess with the final result) and then displays the result in form of average speed stated in both KB/s and MB/s. You can also get source code if you’d like to do some further tinkering with it, or port it. Everything you need to know is on the webpage or available by using -help switch.

Another, more sophisticated, tool (without being too big) is Iperf (a single executable, with source available on the same page). Settings are changed by use of various switches.

For example, the image above shows the port used is changed to 1234, amount of sent data set to 200 MB, interval of reports set to 2 seconds for better accuracy and report format set to MBytes. The usual -help switch brings up further instructions for changing the many additional switches and settings available with this tool

Depending on how advanced the rest of your IT organisation is, you may need to be the jack of all trades. In fact, sometimes it feels as if anything that a Developer or End User doesn’t understand automatically becomes the property of the BASIS Administrator. Typicaly, these can include anything to do with the infrastructure between the users desktop and the SAP application.

An example I’ve been currently working on is a network issue where a user can access the Portal from one machine but not another. I used to use separate tools to do my network monitoring and debugging (yes, there are people responsible for this, but I have won a lot of good will by providing as much data as possible), but these days ….

Net Tools 2008 has been described as The Swiss army tool for network administrators everywhere. It is a very versatile tool, and
just like any tool it can be used for good or evil. What this means is that you may find the site blocked at work.

Available functions, usefull for both the desktop and Windows Servers, include

An FTP Client for quick file transfers,
Monitor system up status with Monitor Host IP,
Mass file renamer, to rename a whole bunch of log files,
Bandwidth Monitor.

Another tool that I’ve feard of is eToolz which is a collection of network and Internet tools that provides a graphical interface for several common commands. This includes ping, tracert, DNS lookups, http headers, default ports, etcetera. This seems more directedt to someone who who supports web sites, but after all thats what the Portal is …

This will be useful for any SAP system running on Windows, or for your own desktop or laptop.

DriveSpacio is a free Windows utility for examining hard drive usage. When you first boot the program you’ll get a list of each hard drive and partition on your system, along with, some details like the files system, cluster size, and a pie chart or bar graph showing used and free space.

But the fun really begins when you click on the Folders section. You can choose a folder, or an entire drive (just click C: or E: or what have you in the browse window), and DriveSpacio will plot your folders on a graph showing you which of your folders are eating the most space. It takes a few moments to scan folders with a lot of subfolders or files, but the result is a pretty effective tool for figuring out why you’re running out of space on your hard drive.

WinDirStat is another free utility which does pretty much the same thing. But if you prefer the bar graph/pie chart look to WinDirStat’s more abstract-looking visualization, DriveSpacio is worth a look. But there is at least one area where WinDirStat holds the edge: while you can delete files from the WinDirStat interface, DriveSpacio only shows you file and folder names. It doesn’t let you launch or delete files.

From Shell Extension City