Kris' Blog

We have been working to update our Weblogic Server 11g (10.3.5) environments to Weblogic 12c (12.1.3). Soon after creating our first 12.1.3 domain, we began seeing our domains AdminServer consuming over 90% of the servers available CPU. After analyzing thread dumps from the AdminServer, we discovered that the threads that were consuming all of the CPU were related to LDAP and our LDAP Provider.

After engaging Oracle Support, a solution was found which has eliminated the High CPU condition. This post just lists the solution that fixed our issue, a future post will outline how we identified the issue was related to LDAP using thread dumps.

The issue relates to the SocketTimeout value being used by Weblogic 12c with respect to LDAP connections. To get around this issue, add the following JVM Parameters to your AdminServer and all Managed Servers which make use of an LDAP Provider:

-Dweblogic.MuxerClass=weblogic.socket.PosixSocketMuxer

More information on this setting can be found here:

http://docs.oracle.com/middleware/1213/wls/PERFM/wls_tuning.htm#BABEBBAH

Recently I encountered a neat problem where I was trying to run a SQLPlus command from within a Shell Script using a schema account which had a special character in the password. The special character was a '$'.

So being a Unix Admin, I knew that chances were the special character needed to be escaped somehow. So here is what I found.

If you try and run SQLPlus directly from the command line with a schema password which contains a special character, you must escape the special character with a '\'

sqlplus user/pa\$sword@XE

Now if you want to run that same command within a Shell Script, you must wrap the password portion in Double Quotes, something like:

Now I know what your going to say, your thinking Oracle WebTier allows you to associate an WT instance with a Weblogic Domain and when you do this Cloud Control will auto-discover it. That is true, but when you are managing large scale environments, sometimes it doesn't make sense to associate WebTier instances with individual domains. This is particularly true when you are running several domains on the same server, why have individual WebTier instances per domain when 1 or 2 will do! 

Having standalone WebTier instances does have a couple of drawbacks with respect to Oracle Cloud Control 12c. The most obvious is that standalone WebTier instances are not automatically discovered like other targets. This post outlines how you can add your standalone WebTier instances as targets to Cloud Control 12c.

To add a standalone Oracle WebTier instance to CC12, you must add it by specifying its target properties. To do this:

Within CC12, navigate to 'Setup' -> 'Add Targets' -> 'Add Targets Manually'

Select the 'Add Non-Host Targets by Specifying Target Monitoring Properties' radio button.

From the 'Target Type' drop down, select 'Oracle HTTP Server'. Then select the agent for the host that the standalone WebTier instance resides on from the 'Monitoring Agent' field and then click 'Add Manually'.

On the next page you will be presented with a long list of property fields. Don't worry, only a few of them are required. Enter the values outlined for the fields below, and for any other required fields just enter 'x'. Once done, click 'Ok'

Machine Name: host name of server WebTier instance resides on

Oracle Home Path: path to the WebTier installation.

Port Number: WebTier listening port (i.e 7777)

Version of the Apache Server: 2.2

There, you've added your standalone WebTier instance to CC12. It will take CC12 a couple of minutes to start aggregating data for the new target so be patient, but once this is complete, you will now be able to see the availability status in CC12.

Recently I upgraded to VMWare Fusion 5 and I started to notice some pretty significant performance degradation on my MacBook. What I was seeing in the 'Activity Monitor' app was after I start up a VM, Fusion would begin consuming just about all free memory on my host. But what was even stranger was that the 'Inactive' memory seemed to continue to climb. It appeared as though Fusion was ignoring the RAM value set for each VM and was just grabbing more and more memory as the VM needed it.

I found this post on the VMWare KB ( http://communities.vmware.com/thread/86878?start=30&tstart=0 ). The post discusses how memory is released depending on how VM's are exited (Suspended or Shutdown). However, one post recommended that the 'Hard Disk Buffering' option be disabled  to see if performance improved. Once I disabled this option on each of my VM's, performance on my MacBook returned to normal. 

VMWare states for this option that "Hard Disk Buffering can increase performance of your Virtual Machine, but also consumes much more memory on your Mac". So my thought is that for VM's which only require a small allocation of RAM to operate, enabling this setting is fine. However, for VM's which require a larger allocation of RAM to operate, the performance gained by the VM is lost because more of the Host's memory is used.

Over the past couple of months I have been tasked on a number of projects to upgrade from Oracle 10g middleware to Oracle's 11g Fusion Middleware. As an IT resource who primarily deals in the middleware space I can say that this is a welcomed initiative!

However, during the upgrade of a number of environments I was running into a strange situation that baffled myself and my colleagues for a few weeks. Today though, we finally sorted it out.

As part of our move to Oracle Weblogic Application Server, we also wanted to place the Oracle WebTier (primarily Oracle HTTP Server) out in front of each install to better manage and secure our web traffic. The installations of both Weblogic Application Server and Oracle WebTier were very straight forward and configuring Oracle HTTP Server to talk to Oracle Weblogic Server was a piece of cake with the mod_wl_ohs plugin.

The issue we ran into though was related to the way that the Weblogic Admin Console redirects traffic after saving a configuration change. We were able to navigate the Admin Console through when accessing through Oracle HTTP Server, but when we made a configuration change, and clicked save we were being redirected away from the Oracle HTTP Server to the Weblogic AdminServer port. Confused yet, me too! Example time:

We were accessing the Admin Console through Oracle HTTP Server so our URL was:

http://localhost:7777/console

While navigating through the console we remained on localhost:7777. However, when we clicked save after making a configuration change, we were redirected to:

http://localhost:7001/console

I know what your thinking, big deal. Well, this does pose a big problem if you plan on accessing the console through a reverse proxy. Because when the redirect happens, you will get a nasty connection reset message because you may not necessarily have access to port 7001.

Today I encountered an issue which when I think about it, has to have one of the strangest solutions yet! So I just had to share.

We have been ramping up for a project at work over the past few months which requires a significant number of servers to implement. Because of this we decided we would create our own sandbox environment so we could start working through some of the low-level environment configuration.

I decided to do this using an old server (Ubuntu Server 64-bit) that was we had kicking around, Oracle VirtualBox and RHEL 5.6. The plan was to create two guest OS' with RHEL 5.6. As always with Ubuntu, installation of VirtualBox took no time at all and I was ready to start creating my first guest VM.

Following the typical steps to setup the VM, I then inserted the RHEL Install cd and started the new VM at which time I was presented with the following error:

verr_vmx_msr_locked_or_disabled

No problem, a quick google of the error informed me that the issue is mainly due to the fact that Hardware Virtualization was not enabled in my BIOS. Sure enough, jumping into the BIOS, I found that Virtualization was not enabled, so I set it to enabled and restarted the Ubuntu Server.

Now back to my VM, I started it back up, and again:

verr_vmx_msr_locked_or_disabled

Then several more attempts with different VM settings and the same result. I began checking some more sites found by google, and finally I came across a post which suggested enabling Hardware Virtualization in the BIOS, saving the changes, and then unplugging the computer from its power source. I have to admit, at first read I was quite skeptical, but after a few more tries with VirtualBox and the same error, what did I have to lose.

Sure enough, back in the BIOS I disabled Hardware Virtualization, and then re-enabled it after which I unplugged the Ubuntu Server from the power-source and then went to grab a coffee. When I returned, I plugged the server back in, started it up, and low and behold, I was now able to do my installation without any issues.

No doubt over the past few months you too have seen the television commercials for the MagicJack claiming to offer unlimited calling within North America. I remember seeing the commercial the first time and thinking, not likely.

Over the past year I have become more and more reliant on my iPhone as a primary means for making calls that having a landline at home seems like a major waste of money. Recently I moved to a new residence and when it came time to consider a landline, I decided to give the MagicJack a try and see if it was all it's hyped up to be. $20 CDN for unlimited calling for one year, worth a shot right.

Well, I must say the MagicJack is everything it claims to be. The installation of the service was a breeze and I have been making calls, both local and long distance since without issue. Installation of the MagicJack took roughly 5 minutes and installation consisted of plugging the MagicJack USB device into a USB port on your computer. Once connected, MagicJack automatically installs the MagicJack call software to your computer and once complete, the software automatically starts up at which point, you are ready to start making calls.

MagicJack does require you to register which also took a total of 5 minutes to complete. During registeration you are offered other features which can be added to your subscription. I added a local phone number (Nova Scotia - 902). Once the registration is completed, MagicJack takes a few minutes to activate your account and then BANG, you are ready to make calls! I can say that from the offset, I am a happy and impressed MagicJack customer.

The thing I was most skeptical about with the MagicJack was call quality. Given that the phone calls are made via VoIP, I expected conversations to experience delays and sound quality to be erratic. That simply is not the case. I have made a dozen calls since setting things up and the call quality has been terrific, with no delays in conversation and great voice quality. But it is not only voice calling you get with the purchase of the MagicJack, you also receive Call Waiting, Caller ID and Voicemail.

There are some things that you do need to consider before jumping on the bandwagon. This service requires that the software included with the MagicJack be running on your computer before you can receive or make phone calls. One nice feature I discovered though is that if the software is not running and someone calls your number, they will be forwarded to voicemail. One other nice feature about this is that you will receive an email anytime a voice message has been left for you as well.

The second thing to consider is that MagicJack does not come with a local phone number for your area. If you want to have a phone number for your area code MagicJack charge an additional $10/year for this service.

Recently, I moved to a new condo complex and one of the first things I had installed was Internet. Like most IT Professionals, I have surrounded myself with devices which get the most from having a WI-FI connection (Playstation, iPhone, MacBook, you get the picture). 

Once Wireless Internet was up and running in my new place, I wanted to make sure that I was getting the best possible connection. Here are some of the steps I took to make sure I was getting the most out of my WI-FI connection.

Most ISP's these days are shipping routers to their customers which have built in Wireless capability. If your like me and have your own Wireless Router, the first thing you will want to do is disable the Wireless Adapter on your ISP's router. Most routers these days have their own Web Interface for Administration. A quick google in most cases will provide information on how to access the admin page and viola, an option to disable the Wireless!

After connecting my Wireless router (Airport Extreme) to the ISP's router, the next step is to find a channel with the least amount of contention. This can be difficult sometimes if say, like me you live in a condo complex with many people in a single building. The trick is to find a channel which has the least amount of devices broadcasting.

There are alot of different utilities out there which you can use to find specific information about nearby WI-FI hotspots. Because I have a MacBook, I used the KisMac application to find the wireless channel which was least used in my building. I then set my Airport Extreme to broadcast on this channel and instantly, KisMac showed that Airport Extremes Signal was the strongest. Perfect!

This post is worth its weight in Gold. Great explanation on how to install Oracle-XE on a 64-bit Ubuntu Server.

If you admin any Oracle Application Server 10g instances, you've no doubt run into the situation where an OC4J container just will not startup due to an issue with a deployed application. When I first started managing OAS 10g instances, when I encountered this situation I dealt with it mostly by removing the suspect OC4J container and rebuilding it from scratch.

After doing a number of times, I figured there had to be a better way to deal with a bad application deploy. Turns out there is, and it has saved me alot of time over the years.

The first step is to stop the OC4J instance where your application is deployed using opmnctl.

$OC4J_HOME/opmn/bin/opmnctl stopproc process-type=<your oc4j instance name>

The next step is to remove your applications directory in your OC4J instance application directory.

rm -r $OC4J_HOME/j2ee/<your oc4j instance name>/applications/<application name>

Next, remove all other deployment artifacts for your application from the OC4J instance application-deployments directory.

rm -r $OC4J_HOME/j2ee/<your oc4j instance name>/application-deployments/<application name>

Thats it for deleting files, now we just need to edit a few configuration files to make OC4J forget about the application entirely. First, edit the $OC4J_HOME/j2ee/<your oc4j instance>/conf/server.xml file for your OC4J container and remove the following element:

<application          name="YourApplication"          path="../applications/<your application name>.ear"          root="yourApplication"          auto-start="true"/>

Finally, we need to remove a similar element in the $OC4J_HOME/j2ee/<your oc4j instance>/conf/default-web-site.xml. Remove the following element:

 <web-app          application="YourApplication"          name="YourApplication-web"          root="yourApplication"          load-on-startup="true"/>

It's hard to believe that given I am in the technology profession, its taken me this long to start a blog, but better late than never. I'm an IT Professional who deals mainly in Application Integration and Security and specializes in Service Oriented Architecture (SOA). But when I'm not at the office I enjoy Music, Cars and all things sports. 

My hope for this blog is to share information about new and emerging technologies in the IT market today as well as best practices for managing current IT Infrastructure and just anything cool that has a technology twist.