I haven’t blogged for over one year. I started a new job in Oct 2010 and that’s kept me occupied. I’m now longer a full time DBA but still heavily involved in databases. I’m doing pre-sales at EMC focused on database backup with EMC’s backup and recovery product line. I help customers understand how they can use EMC’s Data Domain, Networker and Avamar products to backup Oracle, SQL Server, DB2, Sybase, Greenplum and more. Great job, great company to work for, great customers!

For some internal testing, I was installing Oracle 10g on Oracle Enterprise Linux 5.5 and 5.7. Had to do the following:

1. Needed to install several missing rpms, the usual stuff.
2. Ran the 10.2.0.1 installer with the ignoreSysPreReqs flag since the Linux releases were far newer than the installer.
3. When launching the 10.2.0.5 Oracle patch, the runInstaller command bombed without displaying the GUI and complained about a libXp.so.6 error. Google searches revealed a well blazed trail to the work around, which follows.
   • Obtain the xorg-xll-deprecated library.
     ftp://rpmfind.net/linux/fedora/core/4/x86_64/os/Fedora/RPMS/xorg-x11-deprecated-libs-6.8.2-31.i386.rpm. Install the rpm.
   • Prior to launching runInstaller, update the LD_LIBRARY_PATH environment variable to contain /usr/X11R6/lib, e.g.
     $ export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/X11R6/lib

EXPLAIN PLAN FOR
SELECT e.*, d.dept_name
FROM emp e, dept d
WHERE e.deptno = d.deptno
  AND e.ename='benoit';

Rem
Rem Display the plan using the DBMS_XPLAN.DISPLAY() table function
Rem
SET LINESIZE 130
SET PAGESIZE 0
SELECT * FROM table(DBMS_XPLAN.DISPLAY);

Why use it? It provides explain plans formatted much better than the older pre 9i methods of displaying query explain plans. DBMS_XPLAN also provides more information needed for performance tuning. I thought I was pretty slick having this at my finger tips since 9i came out.

What I didn’t realize is that DBMS_XPLAN has been enhanced in 10G and further in 11G.

The comments for this Jonathan Lewis blog post tipped me off to the new functions. They led to Rob van Wijk’s detailed post on the new display_cursor function, as well as another Jonathan Lewis post on the function.

Needless to say, these enhancements have been around for 5 plus years and I’m just getting to them. Humbling!

By the way, the plan table supports keeping data for multiple statements at the same time. Handy if multiple developers need to run explain simultaneously. Here’s how: provide arguments to the dbms_xplan.display function.

-- Begin spooling to file.
spool my_file_name.txt

-- Capture database name and current user.
select * from global_name;
select user from dual;

-- Run query.
explain plan 
SET STATEMENT_ID = 'developer_01_sql_01' 
INTO plan_table for
SELECT e.*, d.dept_name
FROM emp e, dept d
WHERE e.deptno = d.deptno
  AND e.ename='benoit';

SET LINESIZE 130
SET PAGESIZE 0
SELECT * FROM TABLE(DBMS_XPLAN.DISPLAY('PLAN_TABLE', 'developer_01_sql_01','TYPICAL'));

spool off
exit

I logged into the new oracle certification site which gets data from Pearson. I need to contact Oracle to get my prometric tests listed in the same profile. Will do so tomorrow when I get my paper copies of scores and dates in front of me.

Eventually I decided to create a powershell script that iterates through a list of windows hosts and outputs the value of all ORACLE_HOME entries in the registry plus the main PATH variable. A number of machines are likely to have more than one Oracle client. Or they may have multiple values for ORACLE_HOME even though some have been uninstalled. Getting the PATH value will allow me to see which one was installed most recently, since the most recently installed client should be foremost in the PATH.

I’ve posted the powershell script below. It requires an input file and creates an output file that can be opened in Excel. It illustrates using FOREACH loops, invoking the REG QUERY command against remote machines, using IF logic, and reading and writing to a file and some grep style filtering.

# Written with powershell 2.0. Not sure if it works with 1.0.

# File containing list of hosts.
$host_inputfile = "host_list.txt"

# Name of output data file and of log file.
$ofile = "ohome_values.txt"
$olog = "ohome_run.log"

# Initialize counters for hosts and Oracle Homes.
$total_hosts = 0
# Total homes found
$total_homes = 0

# Update log file to say that processing is starting.
echo "Starting new run of get_oracle_client." | out-file $olog -append
$start_time = get-date
echo $start_time | out-file $olog -append

# Start output data file by putting a header row in.
echo "HOSTNAME, KEY_NAME, KEY_TYPE, KEY_VALUE" | out-file $ofile

# Read into a variable the file containing the list of hosts.
$host_list = get-content $host_inputfile

# Iterate through each host in the list.
foreach ($HOST_NAME in $host_list) {

	# Update the user on progress.
	echo "Working on $HOST_NAME..."
	
	# Run a a command to get a list of all values from the registry for
	# ORACLE_HOME. Several strings need to be excluded, such as 
	# ORACLE_HOME_NAME and agent10g.
	
	$OH_RAW = reg query \\$HOST_NAME\HKLM\Software\Oracle /s | select-string "ORACLE_HOME" 
	
	$OH = echo $OH_RAW | sort | get-unique | select-string -notmatch "ORACLE_HOME_KEY", "ORACLE_HOME_NAME", "agent10g"
	
	# Write Oracle Home values to file. Include a column for the host name.
	# Iterate through each line returned from the remote machine.
	foreach ($line in $OH) {
		
		# Initialize output var
	  $out_var = "$HOST_NAME"
		
		$split_line = -split $line
		
		foreach ($word in $split_line) {
			$out_var="$out_var,$word"
		}
		
		# If there was nothing to be processed, then don't write to file.
		if ($out_var -eq "$HOST_NAME,") {
			break
		}
		
		# Increment Oracle Homes counter variable and write values to file.
		$total_homes = $total_homes + 1
		echo $out_var | out-file $ofile -append
		
	}
	
	# Key for getting PATH:
	# HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\Session Manager\Environment
	
	$path_var = reg query "\\$HOST_NAME\HKLM\SYSTEM\CurrentControlSet\Control\Session Manager\Environment"
	$path_var = echo $path_var | select-string "Path" 
	$path_var = echo $path_var | select-string -notmatch "PATHEXT", "TRACE_FORMAT_SEARCH_PATH", "PSModulePath", "Os2LibPath"
	
	# Write path value to file. Include a column for the host name.
	foreach ($line in $path_var) {
		
		# Initialize output var
	  $out_var = "$HOST_NAME"
		
		$split_line = -split $line
		
		foreach ($word in $split_line) {
			$out_var="$out_var,$word"
		}
		
		# If there was nothing to be processed, then don't write to file.
		if ($out_var -eq "$HOST_NAME,") {
			break
		}		
		
		# Write values to file.
		echo $out_var | out-file $ofile -append
	}
	
	# Increment hosts counter variable.
	$total_hosts = $total_hosts + 1

}

$end_time = get-date

# Update screen.
echo "Start time was: $start_time."
echo "End time was: $end_time."
echo "$total_hosts hosts processed." 
echo "$total_homes Oracle Homes found." 

# Append to log file.
echo "Ending run of get_oracle_client." | out-file $olog -append
echo "$total_hosts hosts processed." | out-file $olog -append
echo "$total_homes Oracle Homes found." | out-file $olog -append
echo $end_time | out-file $olog -append

Upgrade steps were simple, the same as previous Apex upgrades. In my environment, I encountered no problems. I’m running database 11.2.0 on Windows 2008 on VMWare with Oracle HTTP server 11.1.0.

The upgrade steps were essentially:
> Verify backups
> unzip apex_4.0.zip file
> stop Oracle HTTP server and Listener
> As SYSDBA, run apexins.sql file. I ran utlrp.sql after that as there were some invalid objects.
> Copy new set of images from apex_4.0 zip file to Oracle HTTP server images directory
> If desired, as SYSDBA, run PL/SQL code to give appropriate privileges to the new Apex 4.0 user
> Restart Oracle HTTP server and listener.

Everything in my main application worked fine, with the exception of the field list drop down on the find button on interactive reports. I’m using IE 7, and the drop down list does not display correctly. The same drop down list displays correctly in Firefox 3.6.3, so I’m guessing it’s a IE 7 issue. I don’t have IE 8 handy to try. The other parts of interactive reports are working as advertised.

Interestingly, interactive reports views that are saved by users no longer are referenced by tabs but from a drop-down list. And reports saved by developers for public use are now called “Default” reports in the drop-down list. Report views created by users appear under the same drop down list as “Private” reports.

While powershell is object based and is tightly plugged into both the Windows OS and .NET, I took the approach of writing a straight-forward script similar to a unix shell script.

To execute a powershell script, you create a text file that ends with the extension PS1, e.g. backup_script.ps1. Here’s how to run the script from the windows command line.

C:\> powershell -file backup_script.ps1

I learned that powershell scripts can have signed security certificates to prevent rogue scripts from causing damage. Powershell can run in different security modes, ranging from restricted (interactive use only, no scripts) to unrestricted, where you get a prompt in order to run a script from the internet. I settled for “RemoteSigned” mode, where scripts downloaded from the internet must be signed, but not ones that you create. Default mode is restricted, so you’ll need to run the following from the powershell prompt to get local scripts to run:

PS C:\> set-executionpolicy remotesigned

Powershell supports two things that help out the unix shell scripter: aliases and pipelines. There are built-in aliases so that you can use familiar commands like ls and echo for the equivalent powershell commands that have much longer names. The powershell commands behave differently with a different set of command switches, but it’s still helpful to have the aliases. The pipeline support allows one to pipe virtually all powershell commands, unlike the meager pipeline support for windows bat commands.

In powershell 2.0 there’s an email command, send-mailmessage, which DBAs will find handy for monitoring and reporting.

The help system is built in to the powershell command line. Typing

PS C:\> help

PS C:\> help specific-command

What’s radically different about powershell from unix style shell languages is that it’s object based. When you pipe a command to another command, you are piping an object, not text. The receiving command can then reference properties of the object that is passed.

One item not supported in powershell: input redirection. This is a mainstay in unix shell scripting. It’s a handy way to keep your sql statements embedded in your shell script and to use variables derived from the shell script directly in the sql without passing them as positional arguments. By input redirection, I mean the following bit of unix scripting:

#!/bin/sh
sqlplus /nolog << EOF
connect $user/$pass@$sid 
select * from table
where f1 = $another_var;
exit
EOF

Here’s a pared down sample of the powershell script I created. This sample does not do anything with the .NET or object based attributes of powershell, with the exception of the line that deletes old files.

It illustrates the following:
> Getting command line arguments
> Formating a date
> String concatenation, done a la unix shell without a concatenation operator
> Assigning variables
> Launching an external command (nothing special, just run it)
> Sending email
> Deleting files older than a certain date

# Get script input params
$ORACLE_SID=$args[0]
$LOG_DIR=$args[1]
$BIN_DIR=$args[2]

# Assemble other variables
$datetime = get-date -format "MMM.dd.yyyy-HH.m.ss"
$RMAN_TYPE=”disk_backup”
$RMAN_LOG = "$LOG_DIR\$ORACLE_SID.$RMAN_TYPE.$datetime.rman.log"

# Start logging to file and then run RMAN command. Echo is an alias for Write-Output.
$newline = "Start Execution of $RMAN_TYPE backup on $ORACLE_SID"
echo $newline | out-file $RMAN_LOG
get-date | out-file $RMAN_LOG -append
rman cmdfile=$BIN_DIR\rman_backup_disk_compressed.rman using $ORACLE_SID | out-file $RMAN_LOG –append

# Send email. Note that cat is an alias for Get-Content.
$body = cat $RMAN_LOG | out-string 
send-mailmessage -to "dba1@firm.com" -from "dba2@firm.com" -body $body -subject "RMAN log output" -smtpserver mailserver

# Remove files older than 8 days.
# Note that ls is an alias for Get-ChildItem and del is an alias for Remove-Item.
# Also, ? is an alias for Where-Object and $_ is a built-in variable 
# to refer to the collection of objects passed from the previous pipeline.
ls $LOG_DIR\$ORACLE_SID.$RMAN_TYPE.* |? {$_.LastWriteTime -lt (get-date).AddDays(-8)} | del

I ran the Oracle Universal Installer (oui) and chose the “Deinstall” button, which is located where it’s always been. But instead of beginning the deinstallation set of screens, it just popped up a message box reading, “Please run the deinstall command $ORACLE_HOME/deinstall/deinstall.”

It turns out there is a new utility just for deinstallation. It is uniform across all Oracle platforms. The documentation details can be found here.

I was sufficiently surprised that I checked some 10.2.0 and 11.1.0 installations. They did not have the $ORACLE_HOME/deinstall directory, so it definitely is new for 11.2.0.

I wrote in my last post that I needed to make a few adjustments to work successfully with a proxy server in my environment. My environment: Apex 3.2.1.00.12, OHS 11.1.0, DB 11.2.0, Win 2008.

All seemed well until I tested pages with flash charts. First off, the IE browser would throw up the message box “This page contains both secure and nonsecure items. Do you want to display the nonsecure items?” After dismissing the message box, the page loaded but the charts would not display. Instead they hung while outputting the message “Loading data. Please wait.”

The first issue that needed sorting out was the https protocol, which was behind the “nonsecure items” message box. In Apex 3.2 there are hard coded references to the protocol at Region Definition – Source. One is

codebase="http://download.macromedia.com/pub/shockwave/cabs/flash/swflash.cab#version=6,0,0,0"

pluginspage="https://www.macromedia.com/go/getflashplayer"

codebase="#HOST_PROTOCOL#://download...etc.

The next issue had to do with the proxy server.

The Region – Source screen also uses #HOST# variable, twice in fact. With the proxy server in place, #HOST# came back with the wrong value. Some surfing on the Apex forum at OTN revealed a dads.conf adjustment, which worked.

PlsqlCGIEnvironmentList        HTTP_HOST=proxyhosturl:port/dir
PlsqlCGIEnvironmentList        REQUEST_PROTOCOL=https

NOTE: don’t include the https:// prefix in the HTTP_HOST value.

Prior to that I had tried a more laborious work around. I created an Application Item called MY_HOST. I set this in an Application Process that runs On New Session – After Authentication. Then in every Region – Source for flash charts, I made the following change:

XMLFile=#HOST#apex_util.flash?

XMLFile=&MY_HOST.apex_util.flash?

That also worked, but involved updating the Source yet again for each chart.

A final point. The charts available in Interactive Charts are also affected by https and proxies. Fortunately, making the above dads.conf adjustment is all that’s needed.

In summary, getting flash charts to work with https and a proxy requires updating the code for each chart so that https doesn’t generate a prompt, plus updating the dads.conf file.

I’ll be interested to try out 4.0 in my shop to see how that works with proxy and https.

In a deployment of Oracle Apex I’m working on, I needed to work with a proxy server. The proxy server has an https url. I’ve also deployed https on my Oracle HTTP server on the Oracle Apex machine.

I’m using the default https certificate that ships with Oracle HTTP Server 11.1.0. While this is not a signed certificate, end users never see a prompt from their browser because this is handled at the proxy server just once by the server administrator.

I’m using the standalone installation of OHS which is now part of the larger Weblogic suite. OHS and Oracle database are on the same Windows 2008 machine. Database version: 11.2.0. Apex Version: 3.2.1.00.12.

Issues:
1. I needed to get a value passed in from the proxy server so that users did not have to login. Here’s how.

2. I wanted to base my security on the username passed in from the proxy. To do that, I created an application item which is in the Shared Components – Logic – Application Items section. Then I created an Application Process (also in the Shared components – Logic) section. This process sets the application item during authentication time.

3. My authentication scheme is database.

4. I updated the authorization scheme to use Application Item mentioned above rather than default built in variable APP_USER.

5. I put in some conditional display of items which detects whether or not a user accessed the application via the proxy server. If they did not, then the Application Item is not set and they will not have sufficient authorization to use the application.

6. Finally, I updated the authorization values accordingly as needed on tabs, lists and pages.