Posts Tagged security
Auditing for Physical Standby Databases with Real-Time Query
Today I got asked by a colleague whether we can do auditing on a Physical Standby that does Real-Time Query. My response was: Probably yes, with AUDIT_TRAIL=OS for the Standby and the audit command specified on the Primary. I did just test that and it really works this way! There is no need to have the AUDIT_TRAIL parameter value different from NONE on the Primary even. Just call a command like
SQL> audit session whenever not successful;
on the Primary and it will reach the Physical Standby via Redo-Apply. I tested that with
SQL> select * from v$version; BANNER -------------------------------------------------------------------------------- Oracle Database 11g Enterprise Edition Release 11.2.0.3.0 - 64bit Production PL/SQL Release 11.2.0.3.0 - Production CORE 11.2.0.3.0 Production TNS for Linux: Version 11.2.0.3.0 - Production NLSRTL Version 11.2.0.3.0 - Production
but I suppose it will work the same with lower versions. But then: “Don’t believe it, test it!”
Fine Grained Auditing and SYS_CONTEXT
For an Oracle Database, we have several ways to audit user activity. With this posting, we will take a closer look at Fine Grained Auditing (FGA), working together with the very useful standard function SYS_CONTEXT. The demonstration is done with 11gR2 but should work very similar with older versions down to 9i, where FGA was introduced, if I recall that right.
For a general overview about auditing, see my presentation slides here.
The standard function SYS_CONTEXT shows us many interesting properties of the connected sessions and is not only useful in the context of auditing:
SQL> connect sys/oracle@orcl as sysdba
Connected.
SQL> select * from v$version;
BANNER
--------------------------------------------------------------------------------
Oracle Database 11g Enterprise Edition Release 11.2.0.2.0 - 64bit Production
PL/SQL Release 11.2.0.2.0 - Production
CORE 11.2.0.2.0 Production
TNS for Linux: Version 11.2.0.2.0 - Production
NLSRTL Version 11.2.0.2.0 - Production
SQL> set serveroutput on
begin
dbms_output.put_line('AUTENTICATION_METHOD: '||sys_context('userenv','authentication_method'));
dbms_output.put_line('CLIENT IDENTIFIER: '||sys_context('userenv','client_identifier'));
dbms_output.put_line('CURRENT_SCHEMA: '||sys_context('userenv','current_schema'));
dbms_output.put_line('CLIENT HOSTNAME: '||sys_context('userenv','host'));
dbms_output.put_line('INSTANCE: '||sys_context('userenv','instance_name'));
dbms_output.put_line('CLIENT IP ADDRESS: '||sys_context('userenv','ip_address'));
dbms_output.put_line('IS SYSDBA: '||sys_context('userenv','isdba'));
dbms_output.put_line('OS USER: '||sys_context('userenv','os_user'));
dbms_output.put_line('SERVER HOSTNAME: '||sys_context('userenv','server_host'));
dbms_output.put_line('USER: '||sys_context('userenv','session_user'));
dbms_output.put_line('SESSION ID: '||sys_context('userenv','sid'));
end;
AUTENTICATION_METHOD: PASSWORD
CLIENT IDENTIFIER:
CURRENT_SCHEMA: SYS
CLIENT HOSTNAME: uhesse-PC
INSTANCE: orcl
CLIENT IP ADDRESS: 192.168.56.10
IS SYSDBA: TRUE
OS USER: uhesse-pc\uhesse
SERVER HOSTNAME: uhesse
USER: SYS
SESSION ID: 33
PL/SQL procedure successfully completed.
Notice that there is a difference between “current_schema” and “session_user”: The first would change after an “alter session set current_schema=someotherschema;” but the second stays the same.
Following creates a small demo schema with one table, containing the sensitive column “BUDGET”:
SQL> grant dba to adam identified by adam; connect adam/adam create table adam.audept (deptno number, dname varchar2(20),budget number); insert into adam.audept values (10,'DELIVERY',10000); insert into adam.audept values (20,'MARKETING',100000); commit;
End-users connect with the Database user ADAM, using the “Audept Application”. Normally, they do not access the audept table directly. Instead, they use a view that does not show the budget column:
SQL> create or replace view audeptvu as select deptno,dname from audept; View created.
Our goal is now to audit access on the audept table, but only if the sensitive column “BUDGET” is touched and even then only if the access is not done through the “Audept Application”. This goal cannot be reached with traditional auditing, because this would audit each & every select on the table:
SQL> connect / as sysdba Connected. SQL> truncate table aud$; Table truncated.
AUD$ is the base table for the Data Dictionary View DBA_AUDIT_TRAIL, like FGA_LOG$ is for DBA_FGA_AUDIT_TRAIL. As sys, we can truncate it, which can be monitored as well. A more sophisticated way to restrict and monitor activities of highly privileged users like sys would be the separate products Database Vault (restrict) resp. Audit vault (monitor).
SQL> audit select on adam.audept by access;
Audit succeeded.
SQL> connect adam/adam
Connected.
SQL> exec dbms_session.set_identifier('AUDEPT_APPLICATION')
PL/SQL procedure successfully completed.
SQL> select * from audeptvu;
DEPTNO DNAME
---------- --------------------
10 DELIVERY
20 MARKETING
SQL> select count(*) from dba_audit_trail where obj_name='AUDEPT';
COUNT(*)
----------
1
Although the access on the audept table was completely compliant with our “Business Rules”, it was audited by traditional auditing. Therefore, FGA was introduced:
SQL> connect adam/adam
Connected.
SQL> create or replace function auditif
return number
as
begin
if
sys_context('userenv','session_user')<>'ADAM'
or
sys_context('userenv','client_identifier')<>'AUDEPT_APPLICATION'
or
sys_context('userenv','client_identifier') is null
then return 1;
else return 0;
end if;
end;
/
Function created.
This function will be used later on to determine the Audit Condition for Fine Grained Auditing. We could specify many more conditions (from SYS_CONTEXT or else) here of course. Although it was tempting (at least for me) to use a boolean return value, it is not a good idea: You cannot test it then with select from dual. Fine Grained Auditing is implemented with the DBMS_FGA standard package:
SQL> begin dbms_fga.add_policy(object_schema=>'ADAM', object_name=>'AUDEPT', policy_name=>'AUDEPT_POLICY', audit_column=>'BUDGET', audit_condition=>'ADAM.AUDITIF=1'); end; / PL/SQL procedure successfully completed.
With this policy, an audit entry is only generated if the column “BUDGET” is touched. Additionally, in order to get audited, the access must be done without using the “Audept Application”. We look at a compliant access first:
SQL> connect / as sysdba Connected. SQL> truncate table fga_log$; Table truncated. SQL> select count(*) from dba_fga_audit_trail; COUNT(*) ---------- 0 SQL> connect adam/adam Connected.
A Client Identifier can be used to differentiate multiple (application or human) users, connecting with the same Database user. This is a common technique, executed (and verified) at the application layer:
SQL> exec dbms_session.set_identifier('AUDEPT_APPLICATION')
PL/SQL procedure successfully completed.
SQL> select * from audeptvu;
DEPTNO DNAME
---------- --------------------
10 DELIVERY
20 MARKETING
SQL> select * from audept;
DEPTNO DNAME BUDGET
---------- -------------------- ----------
10 DELIVERY 10000
20 MARKETING 100000
SQL> select count(*) from dba_fga_audit_trail;
COUNT(*)
----------
0
No audit entry was generated. Now we do not follow the business rules:
SQL> connect adam/adam Connected. SQL> select * from audeptvu; DEPTNO DNAME ---------- -------------------- 10 DELIVERY 20 MARKETING SQL> select count(*) from dba_fga_audit_trail; COUNT(*) ---------- 0 SQL> select * from audept; DEPTNO DNAME BUDGET ---------- -------------------- ---------- 10 DELIVERY 10000 20 MARKETING 100000 SQL> select count(*) from dba_fga_audit_trail; COUNT(*) ---------- 1
The first SELECT did not show the critical column and was therefore not audited, although no Client Identifier was specified. The second select was audited with many details, though:
SQL> select db_user,os_user,object_schema,object_name,policy_name,sql_text,timestamp
from dba_fga_audit_trail;
DB_US OS_USER OBJEC OBJECT_NAM POLICY_NAME SQL_TEXT TIMESTAMP
----- --------------- ----- ---------- -------------------- -------------------- ---------
ADAM oracle ADAM AUDEPT AUDEPT_POLICY select * from audept 22-AUG-11
Notice that FGA will show the SQL text, regardless of the setting of AUDIT_TRAIL – no “EXTENDED” necessary here:
SQL> show parameter audit_trail NAME TYPE VALUE ------------------------------------ ----------- ------------------------------ audit_trail string DB
Conclusion: With Fine Grained Auditing, we have the possibility to limit our audit entries to only non compliant access, according to our business rules. One comfortable way to specify audit conditions is the standard function SYS_CONTEXT.
Added a page about Oracle Database HA Architecture
Because the topic is frequently appearing in my courses, I have added a page to my Blog, dedicated to Oracle Database High Availability Architecture. It talks briefly about the differences between Single Instance, RAC, Extended RAC, Data Guard and Remote Mirroring.
How to audit sys into an OS file owned by root
Sometimes, the requirement may arise to audit all actions of the Oracle Database superuser sys. This feature was introduced already in the 9i version, together with the parameter AUDIT_SYS_OPERATIONS. If set to true, that leads to the generation of files in the directory, determined by the parameter AUDIT_FILE_DEST. These files will contain then a protocol of all the actions of sys. By default, the files in that directory – all with the suffix aud - contain only the protocol of all connections as sys, but not all the actions done after the connect. They look like this:
Audit file /u01/app/oracle/admin/orcl/adump/ora_1057.aud
Oracle Database 10g Enterprise Edition Release 10.2.0.1.0 - Production
With the Partitioning, OLAP and Data Mining options
ORACLE_HOME = /u01/app/oracle/product/10.2.0/db_1
System name: Linux
Node name: uhesse
Release: 2.6.9-55.ELsmp
Version: #1 SMP Fri Apr 20 17:03:35 EDT 2007
Machine: i686
Instance name: orcl
Redo thread mounted by this instance: 0 <none>
Oracle process number: 0
Unix process pid: 1057, image: oracle@uhesse
Tue Feb 2 17:43:12 2010
ACTION : 'CONNECT'
DATABASE USER: '/'
PRIVILEGE : SYSDBA
CLIENT USER: oracle
CLIENT TERMINAL: pts/6
STATUS: 0
But even if we set AUDIT_SYS_OPERATIONS=true, although the audit file will then contain all the actions of sys also, it belongs to the operating system user, used to install the database – usually the user oracle. And likely, the person who has the privilege to connect as sys also has the privilege to connect as oracle on the OS level. That makes the approach with AUDIT_SYS_OPERATIONS somewhat useless from a security viewpoint. That is why we introduced in version 10g (already) the possibility to have the audit file owned by root – which may make it harder to manipulate or delete it for the DBA person. This is how that works:
First, we have to set two parameters – the cooperation of someone able to logon as sysdba is required here…
SQL> select * from v$version; BANNER ---------------------------------------------------------------- Oracle Database 10g Enterprise Edition Release 10.2.0.1.0 - Prod PL/SQL Release 10.2.0.1.0 - Production CORE 10.2.0.1.0 Production TNS for Linux: Version 10.2.0.1.0 - Production NLSRTL Version 10.2.0.1.0 - Production SQL> show parameter audit NAME TYPE VALUE ------------------------------------ ----------- ------------------------------ audit_file_dest string /u01/app/oracle/admin/orcl/adu mp audit_sys_operations boolean FALSE audit_syslog_level string audit_trail string NONE SQL> alter system set audit_sys_operations=true scope=spfile; System altered. SQL> alter system set audit_syslog_level='LOCAL1.WARNING' scope=spfile; System altered.
Then, we logon as root and modify the file /etc/syslog.conf:
-bash-3.00$ su - Password: [root@uhesse ~]# echo local1.warning /var/log/audit.log >> /etc/syslog.conf [root@uhesse ~]# cat /etc/syslog.conf # Log all kernel messages to the console. # Logging much else clutters up the screen. #kern.* /dev/console # Log anything (except mail) of level info or higher. # Don't log private authentication messages! *.info;mail.none;authpriv.none;cron.none /var/log/messages # The authpriv file has restricted access. authpriv.* /var/log/secure # Log all the mail messages in one place. mail.* -/var/log/maillog # Log cron stuff cron.* /var/log/cron # Everybody gets emergency messages *.emerg * # Save news errors of level crit and higher in a special file. uucp,news.crit /var/log/spooler # Save boot messages also to boot.log local7.* /var/log/boot.log local1.warning /var/log/audit.log
After having done that, we need to restart the syslog logger, also as root user:
[root@uhesse ~]# /etc/rc.d/init.d/syslog restart Shutting down kernel logger: [ OK ] Shutting down system logger: [ OK ] Starting system logger: [ OK ] Starting kernel logger: [ OK ]
The last step is to restart the instance, in order to make the previous changes of the non-dynamic parameters active. If afterwards someone connects to the instance as sysdba, all his actions are recorded:
SQL> connect / as sysdba
Connected.
SQL> select * from v$instance;
INSTANCE_NUMBER INSTANCE_NAME
--------------- ----------------
HOST_NAME
----------------------------------------------------------------
VERSION STARTUP_T STATUS PAR THREAD# ARCHIVE LOG_SWITCH_WAIT
----------------- --------- ------------ --- ---------- ------- ---------------
LOGINS SHU DATABASE_STATUS INSTANCE_ROLE ACTIVE_ST BLO
---------- --- ----------------- ------------------ --------- ---
1 orcl
uhesse
10.2.0.1.0 02-FEB-10 OPEN NO 1 STARTED
ALLOWED NO ACTIVE PRIMARY_INSTANCE NORMAL NO
On the OS level, the oracle user can not even see the content of the file, nor modify or delete it:
-bash-3.00$ ls -rtl /var/log/ total 1260 [...] -rw------- 1 root root 34457 Feb 2 17:40 cron -rw------- 1 root root 66395 Feb 2 17:40 boot.log -rw------- 1 root root 377235 Feb 2 17:44 messages -rw------- 1 root root 1410 Feb 2 17:44 audit.log -bash-3.00$ cat /var/log/audit.log cat: /var/log/audit.log: Permission denied -bash-3.00$ whoami oracle
Only root can:
-bash-3.00$ su -
Password:
[root@uhesse ~]# cat /var/log/audit.log | grep instance
Feb 2 17:44:46 uhesse Oracle Audit[1184]:
ACTION : 'select * from v$instance'
DATABASE USER: '/'
PRIVILEGE : SYSDBA
CLIENT USER: oracle
CLIENT TERMINAL: pts/6 STATUS: 0
Of course, the usual aspects of auditing apply in this case also: It is an effort to do (for the system and for people who maintain it), it is space consuming, it may impose performance problems – especially if now large actions are performed by sys, like running catalog.sql or catproc.sql, for example. These actions may lead to very much auditing now…
How to change the size of Online Redologs?
If you deal with Oracle Databases for years on a daily basis like I do, you sometimes forget that things that seem elementary to you ain’t necessarily so for others, especially newbies in the Oracle realm. I noticed that recently when I saw some discussions in OTN about redo logfiles and their sizes.
When I go back in time about 12 years, that’s where I was the first time responsible for an Oracle Database (version 7, but some things never change). I inherited that Database from the previous DBA who did the setup, creating quite small log groups of a couple 100 KBs in size – which was not uncommon these days, because Oracle templates mentioned that size as “medium”, if I recall it right.
I got log switches almost every 10 seconds or faster, but didn’t think about it since nobody complained and everything seemed to run fine. Then I attended an Oracle University course and I learned that this was way too many log switches in that short time and that I should increase the log files. Believe me, that was an exciting thing to do for me those days! I sat there, fingers crossed (mentally, because I had to type in the commands) and did it veeeery carefuly on our production database then.
Some things to consider with changing the size of online redologs:
- You do it for performance reasons – mostly because your original files are too small and you get log switches too often
- It is an Online Operation – no end user will even notice about it – apart from maybe better performance afterwards
- You will not put your Database at risk with it – if you do it right
Let me give you an example:
SQL> select group#,members,status,bytes/1024/1024 as mb from v$log; GROUP# MEMBERS STATUS MB ---------- ---------- ---------------- ---------- 1 1 INACTIVE 50 2 1 INACTIVE 50 3 1 CURRENT 50 SQL> select group#,member from v$logfile; GROUP# MEMBER ---------- ------------------------------------------------------------ 3 /u01/app/oracle/oradata/orcl/redo03.log 2 /u01/app/oracle/oradata/orcl/redo02.log 1 /u01/app/oracle/oradata/orcl/redo01.log
This is my configuration as done by the DBCA. Now I want to get 100m sized logfiles and I want them mirrored for security reason. First, I create new log groups of the desired size:
SQL> alter database add logfile
('/u01/app/oracle/oradata/orcl/redo_g4m1.rdo',
'/u01/app/oracle/oradata/orcl/redo_g4m2.rdo')
size 100m;
Database altered.
SQL> alter database add logfile
('/u01/app/oracle/oradata/orcl/redo_g5m1.rdo',
'/u01/app/oracle/oradata/orcl/redo_g5m2.rdo')
size 100m;
Database altered.
SQL> alter database add logfile
('/u01/app/oracle/oradata/orcl/redo_g6m1.rdo',
'/u01/app/oracle/oradata/orcl/redo_g6m2.rdo')
size 100m;
Database altered.
SQL> select group#,members,status,bytes/1024/1024 as mb from v$log;
GROUP# MEMBERS STATUS MB
---------- ---------- ---------------- ----------
1 1 INACTIVE 50
2 1 INACTIVE 50
3 1 CURRENT 50
4 2 UNUSED 100
5 2 UNUSED 100
6 2 UNUSED 100
6 rows selected.
I like the above naming convention, because I can determine by the name easily what member of what group I see and because the suffix rdo is not so easy misinterpreted as a mere ascii textfile that can be deleted if space is needed – believe me, that has happend many times and made the DBA of that site curse and suffer
Now I am going to drop the small files – with SQL commands first!
SQL> alter database drop logfile group 1; Database altered.
That was easy, wasn’t it? Notice that the OS file is not gone yet – we have to delete them manually (if they are not OMF). All groups that have the status INACTIVE can be dropped that way.
SQL> select group#,members,status,bytes/1024/1024 as mb from v$log; GROUP# MEMBERS STATUS MB ---------- ---------- ---------------- ---------- 2 1 INACTIVE 50 3 1 CURRENT 50 4 2 UNUSED 100 5 2 UNUSED 100 6 2 UNUSED 100 SQL> alter database drop logfile group 2; Database altered.
If the status is CURRENT or ACTIVE, it won’t be possible to drop the group – you can do nothing wrong here!
SQL> alter database drop logfile group 3;
alter database drop logfile group 3
*
ERROR at line 1:
ORA-01623: log 3 is current log for instance orcl (thread 1) - cannot drop
ORA-00312: online log 3 thread 1: '/u01/app/oracle/oradata/orcl/redo03.log'
CURRENT means that LGWR is actually using this group to write the redo protocol into. If a log switch happens, the status can become ACTIVE or INACTIVE:
SQL> alter system switch logfile; System altered. SQL> select group#,members,status,bytes/1024/1024 as mb from v$log; GROUP# MEMBERS STATUS MB ---------- ---------- ---------------- ---------- 3 1 ACTIVE 50 4 2 CURRENT 100 5 2 UNUSED 100 6 2 UNUSED 100
ACTIVE means, the checkpoint associated to the redo protocol in that group is not yet finished. It is in other words still needed for a possible instance recovery and cant be dropped therefore. But we can force the checkpoint:
SQL> alter system checkpoint; System altered.
SQL> select group#,members,status,bytes/1024/1024 as mb from v$log;
GROUP# MEMBERS STATUS MB
---------- ---------- ---------------- ----------
3 1 INACTIVE 50
4 2 CURRENT 100
5 2 UNUSED 100
6 2 UNUSED 100
SQL> alter database drop logfile group 3;
Database altered.
Now I am finished with my SQL commands and I can continue with deleting the files of the small groups from the OS. The instance was OPEN all the time and no user was in any way concerned by this whole action. By the way, I mirrored the log groups to protect my database against really serious problems that would arise otherways, if a whole log group gets lost – for example because someone deletes a file from the OS.
Recent Comments