DOWNLOAD Sample scripts Oracle Database 10g

TAGS linux, scripting, dba, All

An Introduction to Linux Shell Scripting for DBAs
by Casimir Saternos

Learn some basic bash shell scripts for installing, running, and maintaining Oracle databases on Linux.

Published November 2005

About seven years ago, Oracle released the first commercial database on Linux. Since then, Oracle, Red Hat, and Novell/SUSE have been steadily collaborating on changes to Linux kernel as they relate to database and application performance. For that reason, Oracle Database 10g for Linux includes enhancements that are closely related to the operating system. Now more than ever, DBAs need to have knowledge of and experience on this platform to best administer the systems under their watch.

There is a traditional division of responsibilities between sysadmins and DBAs. However, in practice, the distinction is not always clear. Many IT shops employ individuals who address concerns at the database as well as the operating system levels. And of course the Oracle Database itself uses operating system resources and is designed to interact closely with its environment.

Furthermore, many sysadmins and DBAs find it necessary or convenient to automate tasks related to their work. The installation of software, monitoring of system resources, and management of systems involve repetitive and error-prone tasks are better addressed through automated processes than manual procedures.

One method that is used to automate such tasks is shell scripting. Shell scripts play a significant role in the Linux System from the time it is installed. Various scripts are called when the system is started up and shut down. Utilities by Oracle and other third-party vendors are invoked through shell scripts. Because they can be developed quickly, they have historically been used for prototyping applications. System Administrators have taken advantage of the functionality available through shell scripting to provide solutions that are tailored for the particular requirements and idiosyncrasies of the systems under their watch.

In this article, I will introduce functionality available through "bash" shell scripting that is relevant to installing, running, and maintaining Oracle databases on a Linux platform. Note that this article is designed for Linux scripting beginners or DBAs who are relatively new to Linux; most experienced Linux sysadmins will not find it helpful.

What Is a Shell Script?

A shell script is simply a text file containing a sequence of commands. When you run the file—or script—it executes the commands contained in the file. The term shell simply refers to the particular command-line user interface you use to communicate with the Linux kernel. Several different shells are available, including the C shell (csh), Korn shell (ksh), Bourne shell (sh), and Bourne-Again shell (bash). The shell itself is a command that reads lines from either a file or the terminal, interprets them, and generally executes other commands. The Bourne-Again shell incorporates features of the other shells mentioned and is the one that was used for the scripts this article demonstrates.

The first line in the script file can be used to dictate which shell will be used to run the script. This is the meaning of the first line found in all of the script examples:

#!/bin/bash
	

Depending on your background, you may not see any immediate value to shell scripting as it relates to the DBA’s work. If you do not have experience with UNIX or UNIX-like systems, the myriad of cryptic commands might be cause for concern. Besides, in addition to being a relational database, Oracle 10g provides a robust platform for processing data within the database as well as several methods of interacting with the OS outside of the database.

However, there are several reasons you might find yourself delving into the world of shell scripting, including the following:

• You are in a situation in which you must support already existing scripts.

• You want to automate system setup that occurs prior to the installation of the Oracle software. For instance, you could write a script to check the initial state of the OS and report any prerequisites that must be met before installation of the software. The script might also create relevant OS users and groups and set environmental variables for the users.

• A running Oracle database can be used to execute manual or scheduled tasks. However, some tasks need to be run when the database is not running. You can use a script to stop or start a database (as well as a listener or a related database process). Such an action cannot be initiated from within the database itself.

• You need a mechanism for monitoring the state of a database (i.e. if it is running and available to process queries). Such a script could also monitor other processes and resources that are not Oracle-specific to provide a fuller picture of what is occurring in the system.

• You need to automate backups. Oracle Recovery Manager (RMAN) is a utility that allows you to develop backup scripts that can run on any platform. You can call Oracle Recovery Manager from a shell script and use it to perform a wide range of backup and recovery activities.

• You might have a requirement that is not specific to a single database. Perhaps you have several databases on a single machine. It may not be advisable to fulfill the requirement by using a single database, and it introduces potential security issues as well. Under such circumstances, shell scripting provides a means for fulfilling the requirement in a manner that does not associate the process with a single database.

Oracle Database includes functionality that extends beyond the traditional definition of an RDBMS. Like any other piece of software, it uses resources that are provided by the operating system, but it can “see” and “change” its environment to a much greater degree than other software. SQL and Oracle’s fixed views provide a picture of the system from inside the database, whereas shell scripting provides a view of the system from outside of the database. Shell scripting is not the solution for every problem.

It important to recognize that many aspects of the operating system can be monitored and modified from within the database. Oracle’s fixed views (the views with a v$ prefix) can be used to determine the host name of the machine (v$instance) or the platform name on which the database is running (v$database). The location and other attributes of the files associated with the database can be determined in this manner as well. The location and other attributes of datafiles (v$datafile, dba_data_files), temp files (v$tempfile, dba_temp_files), redo logs (v$logfile), archive logs (v$archived_log), and control files (v$controlfile) can be queried directly from the database. You can determine information about the flash recovery area ($recovery_file_dest) from this view, as well as by looking at some init.ora parameters (db_recovery_file_dest, db_recovery_file_dest_size). The status of processes (v$process) and memory (v$sga, v$sgastat, and so on) can be queried as well. There are various built-in PL/SQL packages as well as the ability to create Java and C database objects that allow for additional access to the underlying OS.

If you are considering scripting for a task that requires a good deal of database access, scripting is probably not your best option. Later in this article, there is a description of how to access the database by using SQL*Plus but in many cases, you are better off approaching the problem by using another language.

The charts below summarize information accessible from within the database itself:

BASH Scripts

Scripts are either called as part of an automated process (with no human intervention) or run interactively, with a user responding to prompts. As long as you have executable permission for a file, you can run it from the command line by typing its name. If you do not have executable permission for the file but do have read permission, you can run the script by preceding it with sh.

If a script is designed to be run without user input, several options are available for calling it. You can run a script in the background and continue even if you disconnect, by entering a command in the following form:

nohup /path_to_dir/myscript_here.sh &

The following examples cover the essential aspects of providing output to view (using echo), looping, conditional logic, and variable assignment.

print_args.sh. Arguments are words to the right of the command name that are passed into the script. To access the first parameter, you use the $1 variable. The $0 variable contains the name of the script itself. The $# variable contains the number of arguments in the script. A handy way to iterate through all of the parameters passed involves the use of a while loop and the shift command. This command is what lets you iterate through all the arguments in the argument list (rather than remaining in an infinite loop).

while [ $# -ne 0  ]
do
        echo $1
        shift
done

if [ ! -r $1 ]; then # not exists and is readable
        		echo "File $1 does not exist or is not readable."
       		exit;
fi	

	
if [ ! -r $1 ];

do_continue.sh. This example is a simple representative sequence of commands that can be used to read user input for various purposes. Before running a process that can result in data loss or other undesirable results under certain conditions that are not determinable from within the script, it is advisable to include a prompt asking if the user actually wants the script to execute the next command or commands. The following example asks if the user wants to continue, reads a variable named doContinue from the command line, and evaluates what the user entered. If the user enters anything other than “y,” that person is informed that the script is “quitting” and it exits without executing the remainder of the script following the end of the if block (fi).

doContinue=n
echo -n "Do you really want to continue? (y/n) " 
read doContinue

if [ "$doContinue" != "y" ]; then
      echo "Quitting..."
      exit
fi

	
echo "You are logged in as ‘whoami‘";

if [ ‘whoami‘ != "oracle" ]; then
  echo "Must be logged on as oracle to run this script."
  exit
fi

echo "Running script at ‘date‘"

	
stty -echo    
	echo -n "Enter the database system password:  "
	read pw
stty echo

Some files reside in a fixed location for a given Oracle install. You can determine the Oracle inventory by viewing the /etc/oraInst.loc file. The /etc/oratab file identifies databases (and other Oracle programs) installed on the server.

get_inv_location.sh. This script is a bit less intuitive than the previous examples. By breaking down the script into its component commands, you will get a better understanding of what is being accomplished.

To determine inventory location, you are going to pipe the results of the cat command (which displays the contents of the file) to grep (a utility that prints lines that match a given pattern). You are searching for lines that contain the literal inventory_loc .

cat /etc/oraInst.loc | grep inventory_loc

	
cat /etc/oraInst.loc |grep -v "#"|grep inventory_loc

The result of this command will look something like this:

inventory_loc=/u01/oraInventory

	
cat /etc/oraInst.loc|grep -v "#"|grep inventory_loc > tmp

cat tmp | awk -F= '{print $2}'

rm tmp

	
$ORACLE_SID:$ORACLE_HOME:<N|Y>:

cat /etc/oratab | awk -F: '{if ($1=="TESTDB") print $2 }'

However, what if you have a requirement that an operation needs to be performed on each ORACLE_HOME listed in the /etc/orainst file? You can iterate through such a list by utilizing the following code snippet.

dblist=‘cat /etc/oratab | grep -v "#" | awk -F: '{print $2 }'‘

for ohome in $dblist ; do
  echo $ohome
done

search_log.sh. A variety of logs are generated by Oracle products, and you might be interested in monitoring them. The database alert log contains messages that are critical to database operations. Log files are also generated when products are installed or deinstalled and when patches are applied. The following script iterates over a file passed to it as an argument. If any lines are found that contain ORA-, an e-mail message is sent to a designated recipient.

	
cat $1 | grep ORA- > alert.err

if [ ‘cat alert.err|wc -l‘ -gt 0 ]
then
        mail -s "$0 $1 Errors" administrator@yourcompany.com < alert.err
fi	

Environmental variables such as ORACLE_HOME, ORACLE_BASE, and ORACLE_SID can be used to locate resources that are not in a fixed location in the Linux environment. If you are administering an Oracle E-Business Suite 11i application instance, you have numerous other environmental variables that can be used to locate resources. These include APPL_TOP, TWO_TASK, CONTEXT_NAME, and CONTEXT_FILE, to name a few. To see a complete list in your environment, execute the following command and examine the resulting file (myenv.txt):

env > myenv.txt

	
$ORACLE_BASE/admin/$ORACLE_SID/bdump/alert_$ORACLE_SID.log

Based on the principles introduced in this script, a larger one can be written and scheduled to execute at periodic intervals that will search the contents of the alert log (or another file of interest) and send an e-mail if any errors exist. Then the contents of the log can be moved to another file, so that only the most recent error messages will be sent via e-mail.

Oracle Recovery Manager Scripts. Oracle Recovery Manager (RMAN) is a utility that can be used to administer database backup and recovery. This greatly simplifies administration on multiple platforms, because all backup scripts can be written to be run by RMAN, reducing the amount of platform-specific code. RMAN can be called by the underlying operating system and passed a script. For example, a cold (cold.sh) backup might consist of the following script:

	
#!/bin/bash
rman target / <<EOF
shutdown immediate;
startup mount;
backup spfile;
backup database;
alter database open;
delete noprompt obsolete;
quit;
EOF	

Nightly backups are commonly scheduled and run automatically. The script above could be called and the contents of standard out sent to an e-mail address with the following command:

sh cold.sh | mail -s"Backup ‘date‘" administrator@yourcompany.com

Other Oracle utilities can be run from within shell scripts as well. The tnsping utility can be used to see if a given Oracle connection identifier can contact the listener. You might run this utility to check for connection problems:

tnsping ptch04 |grep TNS- 

Database Installation. Many of the steps involved in the setup of a database can be automated. Before you install Oracle 10g on Linux, you need to run various tests to verify the minimum required version of packages and the settings of kernel parameters. You can query the version of a package by using the rpm command with the –q option.

	
rpm -q compat-libstdc++

You can determine various aspects of the system by looking at the /proc “virtual” or “pseudo” file system. It contains not real files but, rather, runtime system information that can be viewed as if it resided in files. For instance, /proc/meminfo contains memory information for the system, and grep MemTotal /proc/meminfo displays the total memory of the system. By using awk as you did earlier, you could isolate the amount of memory in kilobytes, by using:

grep MemTotal /proc/meminfo | awk '{print $2}'

Database Monitoring. The ps command can be used to report process status and to check to see if a database, listener, script, or any other process of interest is running. If you want to list all the databases that are currently running on a server, you can run the following command:

	
echo "‘ps -ef | grep smon|grep -v grep|awk '{print $8}'| awk -F \"_\" 
'{print$3}'‘"

exec_sql.sh. As mentioned earlier, it is possible to query the database from a shell script provided the user has access to sqlplus. The following example returns a (space-delimited) list of machines that are currently maintaining sessions in the database:

	
#!/bin/bash

output=‘sqlplus -s "/ as sysdba" <<EOF
       set heading off feedback off verify off
       select distinct machine from v\\$session;
       exit
EOF
‘

echo $output	

Conclusion

Shell scripting can be an effective tool for quickly automating repetitive and error-prone administration tasks. The examples in this article provide an introduction to the possibilities available but are far from comprehensive. Every system has distinct quirks and foibles and a unique configuration. An administrator will develop unique solutions to meet the needs of the particular system.

Send us your comments