Andreas' Adventures

Haven't blogged this before. A while ago I posted that my Oracle VM was not working and that I shut it down. Shortly after this I found that the reason was the faulty disk I was using. Rebuild Oracle VM and still happy about it.

When patching the Oracle OIM 10.1.4 to OIM 10.1.4.3 the first configuration assistant might fail immediately. Try starting the listener and issue a retry. This helped in my case.

Very often when you are busy with VMWare, you need an extra disk. Now creating the disk can happen while the VM Linux guest is active. However you don't see the disk immediately. A reboot will help obviously, but I found a nice script on http://www.garloff.de/kurt/linux/ that will show the new SCSI device, so that you can continue and add the disk without a reboot. Just search for the rescan-scsi-bus.sh script. It is on my VMWare Server now and will go directly to each new Linux guest from now on. Any possibility to prevent a reboot is welcome ;-)

Hmm, sometimes life is funny. If I say to you that I have a generic patch, you would expect one patch that can be applied to all versions. Not in this case. I need to apply a patch for the Oracle IDM Suite (10.1.4). Finding the patch is already difficult - and then there are four versions of the same patch - apparently based on different versions of the underlying J2EE. In my opinion this is not generic. Trial and error led me to one of the four versions - which then came back with: OPatch detects your platform as 46 while this patch xxxxxxx supports platforms: 0 (Generic Platform) So again - it seems that my definition of generic differs from Oracle's view. I would understand an error, if the patch was for platform 46 or 18 or whatever. But here a generic patch is apparently not applicable for my platform. Luckily Metalink Note 404473.1 offers a solution. So setting the following: export OPATCH_PLATFORM_ID=0 does trick the opatch into believing that it is no Linux machine any l...

Today I decided to abandon Oracle VM and move back to VMWare. A while ago I bought my old 4GB laptop from my company, when I was given a new one. I added a 1TB disk and installed Oracle VM 2.2 on it. There are some quirks with Oracle VM when it comes to support low end systems. First of all the system is memory bound quite fast. See an earlier post of me (http://achatzia.blogspot.com/2010/05/oracle-vm-memory-issue.html) to see how Oracle VM treats it memory. The second main issue is the problem with external USB2.0 devices. I ended up rebooting an image quite often as the SCSI drive was declared dead. I worked with VMWare since about 6 years, and only had lost one image - a colleague disconnected my external drive which resulted in a corrupt Oracle database. With Oracle VM I never had that stability. So for me it is good-bye OVM, and welcome back VMWare Server. Obviously the VMWare Server will run on a Linux host with Oracle Enterprise Linux 5.5 .

I was reinstalling my Oracle VM (again). During the configuration of the Oracle VM Manager template I ran into an error: OVM-2007. The template (new OVM image) could not find the host it was running on. Quick check in the /etc/hosts file of the OVM Server showed: 127.0.0.1 ovm.mydomain.local ovm localhost.localdomain localhost So I changed it to read: 127.0.0.1 localhost.localdomain localhost 10.1.1.100 ovm.mydomain.local ovm localhost.localdomain localhost To me as a Unix veteran it is still a mistery why somebody came up with the idea to put anything besides 127.0.0.1 localhost into the first line.

Recently I had a lot of disk errors on my Oracle VM server. The errors indicated that some scsi problems were manifesting itself with: unable to enumerate USB device on port 2 After unloading and reloading the ehci_hcd it seems to be gone.

Today I tried to cut some corners and use a silent.xml file for WLS 10.3.3 for a silent installation of 10.3.2. To my amazement on a clean machine I received the following error: oracle@xxx001:/opt/oracle$ java -Djava.io.tmpdir=/opt/oracle/tmp -jar /nfsstage/wls1032_generic.jar -mode=silent -silent_xml=./silent/wls_silent.xml -log=/tmp/wls.log Extracting 0%....................................................................................................100% The local BEA product registry is corrupted. Please select another Middleware Home or contact Oracle Support I checked the wls_silent.xml file and found that it includes Coherence, which was not bundled with the 10.3.2 version. So I modified the 10.3.3 silent_xml file from value="WebLogic Server/Core Application Server|WebLogic Server/Administration Console|WebLogic Server/Configuration Wizard and Upgrade Framework|WebLogic Server/Web 2.0 HTTP Pub-Sub Server|WebLogic Server/WebLogic JDBC Drivers|WebLogic Server/Third ...

Nobody likes to write documentation, and I'm not that different. But today I was very happy and said "Thank You" to myself. A while ago I needed to install some software for a project. Now after my vacation, which apparently wiped a lot of my mind's hard disk clean, I needed to repeat the installation for a new environment. Reverting back to my own documentation saved me a lot of time. Maybe I should create a video on YouTube with my impersonation of Steven Balmer's monkey dance and substitute the words "Developers, developers..." with "Document! Document! Document!". So take it from me: writing documentation before your vacation keeps you longer in your post-vacation mode ;-)

A few posts ago I mentioned that the silent install was very slow. Still had this on my to-blog list, so here is the clarification. It turned out that somebody had the idea that we needed some kind of virus protection on the Linux boxes. Each time something was written to the disk the virus protection kicked in. Imagine what this is doing to your installation performance :-(

When using the Oracle Database for the storage of WebLogic Portal settings you need to modify the database.properties and run the create_db.sh script. In the database.properties the example uses the Pointbase, so the values in the Oacle section are filled with placeholders. When replacing the jdbc string be carefull not to remove all @ signs. The original value is: oracle.url=jdbc:oracle:thin:@@DB_HOST@:@DB_PORT@:@DB_NAME@ The fisrt @ after the 'thin:@' needs to remain there. Otherwise you will get a sql exception when you run the create_db.sh script.

It seems that I'm all into silent installs these days. Trying to do a WebLogic installation in silent mode. Went to the documentation, as a few things have changed since the early BEA days. So I found the sample file for the silent.xml . Without big thinking I copied the sample like this: &ltbea-installer&gt &ltinput-fields&gt &ltdata-value value="D:\Oracle\Middleware_Home" name="BEAHOME"&gt &ltdata-value value="D:\Oracle\Middleware_Home\wlserver_10.3" name="WLS_INSTALL_DIR"&gt &ltdata-value value="WebLogic Server/Core Application ServerWebLogic Server /Administration ConsoleWebLogic Server/Configuration Wizard and Upgrade FrameworkWebLogic Server/Web 2.0 HTTP Pub-Sub ServerWebLogic Server/WebLogic JDBC DriversWebLogic Server/Third Party JDBC DriversWebLogic Server /WebLogic Server ClientsWebLogic Server/WebLogic Web Server Plugins WebLogic Server/UDDI and Xquery SupportWebLogic Server/Server Example...

Yesterday I did my first ever silent install. Nothing fancy yet. There are two possibilities: a) Silent install is dead slow or b) There is something wrong with the machine or the storage Bottom line is, it took me 1.5 hours to just install the database software. I need to investigate this, as this is not what I had in mind. Continuing now with the listener and the dbca in a silent manner.

... or Google Chrome has a problem with counting ;-)

I finally have made the jump over to Australia. After a somewhat bumpy first week (oh man - what a jet-lag) I started to settle down. On my project I will do a lot of stuff with Coherence, WebLogic Portal, the OSB, Grid Control and a RAC database. I'm looking forward to do this. This is part of re-inventing myself, while being in a nice place. I will also try to blog some of my experiences in http://yetanotheraustraliablog.blogspot.com/

After some first baby steps with Oracle VM I wanted to push this thing a little bit further. So my idea was to have more than a couple of VM's running simultaneously. The Oracle VM told me however that I had not enough memory to do this. So I was sitting in front of my dedicated laptop and told the screen that there should be 4 GB inside. The laptop however had one of his funny days and told me (using free and top) that really nothing more than 572 MB would be inside the machine. Hmm - when a machine tells me something I am the first to listen. So I shutdown everything, got my screwdriver, opened the memory compartment and had a look. Indeed 1 bank of 2GB. I didn't check the second bank as this is hidden underneath the keyboard. Closing the machine and powering it up, I prepared myself for another round of telling it that it should recognize a little bit more than 572 MB. I examined the dmesg and still it told me that there was only 572 MB inside. Now imagine, you have just see...

Another thing I came across when dealing with the OSB was that the Workshop throw an error and didn't start at all. When you end up with a large dialog box showing a list of startup parameters (which all originate from the workshop.ini file) have a look at this site: http://wiki.oracle.com/page/How+to+get+Oracle+Workshop+for+Weblogic+10GR3+working+in+Linux?t=anon

During the startup of the Eclipse Workshop in my OSB environment I encountered the following error: “requested array is larger than heap” The solution (from the OTN forums) is to delete the publish*.dat from the following directory: FMW_HOME/user_projects/workspaces/YOUR_WORKSPACE/.metadata/.plugins/org.eclipse.wst.server.core/publish/

I am busy installing the latest version of the Oracle Enterprise Manager. Obviously I need a database for this. In order to protect my resources in the Oracle VM image that use I switched off all extra Database options. Among others the Partitioning option is offered. Well - during the installation of the OEM I found out the hard way that the Partitioning option is required for the database. I know that the OEM creates a lot of partitions - which will then get emptied during the data consolidation process. I happened to forget this. The funny thing is that the OEM installer checks this and comes back with a warning box, stating: Using the Partitoning option is strongly recommended. Being a wise guy I tried to ignore this. But when Oracle says that they STRONGLY RECOMMEND something they mean it. So I am deinstalling and reinstalling the database software as I write this ;-)

Just for your convenience: When you start an Oracle VM image from the commandline with xm create -c ./vm.cfg you will see that the boot process continues until the login is displayed: EXT3-fs: mounted filesystem with ordered data mode. type=1403 audit(1272761495.554:2): policy loaded auid=4294967295 ses=4294967295 Enterprise Linux Enterprise Linux Server release 5.4 (Carthage) Kernel 2.6.18-164.0.0.0.1.el5xen on an i686 soa.mydomain.local login: In order to get back to your commandline type CTRL-] and you will get back the control.

Yesterday I received a farewell email from my dear colleague David Calcano. After years with Accenture he is moving into a total different area, leaving us with all the mess of Oracle and SOA. I always admired folks who know more than I do, and the more they know more than me the more I dream of becoming like them. And then they turn out to be brave enough and do something completely different. I will keep www.nosecodeproductions.com on my radar - and I know that with people like David on board they can only succeed! And now back to the scared and meek folks who just keep on doing the same - me ;-)

Although I am busy moving to Oracle VM I needed to access one of my systems on VMWare on my laptop which I had not used for a while. The console of VMWare did not load. Ok - I though his might be due to some networks which I had disabled on Windows. Enabled them - no result. Checking the VMWare services I found out that the VMWare Host Agent was not running. So I restarted it. Crashed again. Obviously the next step is to check the Windows event viewer. To my surprise the System Log could not be displayed. Needless to say I had to fix this first. Clear the Log typically solves this and I was lucky this time. Back to the VMWare Host Agent. Started it agian - still crashes. I found the following error: The VMware Host Agent service terminated with service-specific error 4294967295 (0xFFFFFFFF). Checking the VMWare log. Log for VMware Server, pid=7444, version=2.0.2, build=build-203138, option=Release, section=2 [2010-04-24 12:20:34.264 'App' 2600 info] Current working directory: C...

I was worried as I ran into a problem with the Database Vault security. I configured a new realm in a database where the Oracle E-Business Suite is running. The main idea was to have a user for the OBIEE that should have only limited access. Together with some colleagues we made sure that the OBIEE user had the correct grants on some tables and views inside the EBS schema (APPS, AR, GMS, PA). I created the realm and added the OBIEE user as a participant. Everything seemed to work. Then I disabled the realm and even removed the OBIEE user from the realm, but still Oracle Answers was capable of accessing the APPS schema objects. Took me a day and finally I reread the manual. I understood my error. The purpose of the DBV is to lockout users with system privileges, such as the SYS or SYSTEM user. Now the account of the OBIEE user does not have these system privileges except CREATE SESSION. In order to access any object the OBIEE user relies on discretionary grants (e.g. the AR user grants ...

No - I'm not going to be married (been there - did it - bought the ring - still happy). This was a day in which all extremes came together. Talked about Oracle Complex Event Processing (still need to learn something in there), while I was helping some colleagues to solve some Oracle AS 10.1.2 issues. And in between I was investigating some EBS with Database Vault and TDE after a morning session about the strategy for our Oracle unit. For me this proves that even if you teach an old dog some new tricks, there is always an audience for the previous show, while still performing the bread and butter business today.

I need to install an Oracle IDM on a 64bit Linux. In order to get this running I need a 64bit WebLogic Server. As this is not delivered with an included JDK I downloaded the Sun 1.6 JDK for 64bit Linux systems. When I then tried to install the WLS I always received an "out of disk" error. The magic option to pass to the self-extracting WLS is the following: java -Djava.io.tmpdir=/apps/tmp -jar ./wls1032_generic.jar

I have just added three submisions for OOW 2010. Automate the creation of extra Coherence machines based on Grid Control data Use Oracle VM to learn Oracle technology Build a Coherence cluster using Oracle VM to speed up Oracle FMW

One size does not fit all. The Oracle VM template comes with just the twm windows manager. Since my days in University I dislike this thing. For me it is Motif. So in order to get my favorite Windows manager up and running I need to install it. Luckily Oracle offers a public yum server. Do the following: cd /etc/yum.repos.d wget http://public-yum.oracle.com/public-yum-el5.repo Edit this file and set the stanza to enabled for the corresponding entry. The template is based on EL5 Update4. Then just run yum install openmotif Afterwards I modify the xstart file in the /home/oracle/.vnc directory to use mwm instead of twm. In order to access the vncviewer I also need to modify the iptables. Add a line into the iptables to enable access in the port range 5900-5950. This will enable the access of the custom vncserver, as well as the vncserver which runs on the VM Server. The second vncserver acts as a console for the VM image. Depending on the order in which the VM's are started the vnc p...

I could change the hostname of the VM template for the WebCenter. This would be done on the OS using the standard Linux way (change /etc/hosts, /etc/sysconfig/network). However as the Oracle Application Server OC4J is still configured with the template hostname (haovm007) I would need to changethis. Now - as a lot of you know - the hostname and domainname change in the OC4J is not the easiest tasks of the Application Server Administration. So I decided to build a WebCenter machine from scratch. A) It is easier B) I learn all the steps (probably have to install it more than once) C) I configure it as I would like to have it D) It is more fun

The beauty of the Oracle VM templates is of course the ease with which you create a new machine. But then it has to work, or at least it must be customizable. I have created a new machine from a WebCenter VM template. During the boot it asks for the network configuration. This seems to work, as ifconfig returns the correct settings. However the machinename is not used. I entered webcenter1 but the template still has the haovm007: While I am a Bond fan I'd rather see my own name for the machine. Let's see how this can be corrected.

I thought it worked - but of course I forgot the Oracle VM Manager on my second machine. It still displayed the ServerPool1 that I had created previously. In order to begin with a clean slate you might need to drop the database in the Oracle VM Server. Login to the server as root and execute the following commands: service stop ovs-agent rm -rf /etc/ovs-agent/db service start ovs-agent Then the external disk needs to be added again: [root@ovm utils]# ./repos.py --new /dev/sdb1 [ NEW ] 6ad759c5-d706-4b3a-9d14-bcffafc61fc3 => /dev/sdb1 [root@ovm utils]# ./repos.py --list [ ] 6ad759c5-d706-4b3a-9d14-bcffafc61fc3 => /dev/sdb1 [root@ovm utils]# ./repos.py --root 6ad759c5-d706-4b3a-9d14-bcffafc61fc3 [ R ] 6ad759c5-d706-4b3a-9d14-bcffafc61fc3 => /dev/sdb1 The filesystems were back again and also the directories in /OVS. Now back to importing a VM Template.

I rebuild my OVM server for the n-th time this morning. Problem was that it was difficult to have the external disk mounted on the /OVS directory. I tried to add a shared virtual disk - which I formatted as an ocfs2 disk to the existing disk. I found this excellent posting http://geertdepaep.wordpress.com/2008/06/09/oracle-vm-and-multiple-local-disks/ that described the process and guided me into the correct direction. Although the disk was visible in the repos.py script I was unable to see it in the Oracle VM Manager. Now I thought that I need to build the server and format the filesystems myself. Sorry I have no screenshots here as the installer runs on the machine itself. During the installation the two disks (internal 50 GB and external 1 TB) are shown as /dev/sda and /dev/sdb. In the disk setup utility in the installer choose the option to define a custom layout. The installer offers a predefined layout, which it would normally apply to the available disks. I decided to have three...

Although my Oracle VM Manager was running on OEL, I reinstalled the machine with Fedora and obviously I wanted the OVM Manager back on it. The installation was pretty straight forward. Two tiny things had to be done in order to convince the installer to proceed. I needed to comment out the checks for the OS and the libaio package. After this the installer was on its way. During the Oracle XE installation I adapted the swapfile (created a big additional swapfile) as the XE installer decided to stop when tere was not enough swap space. Conclusion: OVM Manager runs fine on Fedora.

Added a 1 TB disk to my OVM machine :-) [root@ovm ~]# mkfs -t ext3 /dev/sdb mke2fs 1.39 (29-May-2006) /dev/sdb is entire device, not just one partition! Proceed anyway? (y,n) y Filesystem label= OS type: Linux Block size=4096 (log=2) Fragment size=4096 (log=2) 122109952 inodes, 244190646 blocks 12209532 blocks (5.00%) reserved for the super user First data block=0 Maximum filesystem blocks=0 7453 block groups 32768 blocks per group, 32768 fragments per group 16384 inodes per group Superblock backups stored on blocks: 32768, 98304, 163840, 229376, 294912, 819200, 884736, 1605632, 2654208, 4096000, 7962624, 11239424, 20480000, 23887872, 71663616, 78675968, 102400000, 214990848 Writing inode tables: done Creating journal (32768 blocks): done Writing superblocks and filesystem accounting information: done This filesystem will be automatically checked every 38 mounts or 180 days, whichever comes first. Use tune2fs -c or -i to override. [root@ovm ~]# So - now the fun...

Sometimes error messages are very misleading. If you see an error stating: Exception in thread "main" you expect that this is some java-related issue. Maybe the classpath is wrong, your JAVA_HOME is not set, etc. I had this error when I was starting up the Oracle dbca. After some checks and different tries I found out that the DISPLAY environment was set, but to a different screen, which I was not using. So even when you check the dbca shell script, you will find a build-in check for the DISPLAY environment. But now the check passed with flying flags - as it was set. The problem that my dbca was unable to use this DISPLAY lead to the java error. Setting the DISPLAY to the one I was using (and setting xhost to allow all) solved this.