8i | 9i | 10g | 11g | 12c | 13c | 18c | 19c | 21c | 23ai | Misc | PL/SQL | SQL | RAC | WebLogic | Linux
Liquibase : Automating Your SQL and PL/SQL Deployments
This article gives a quick overview of applying changes to the database using Liquibase.
There is a rewrite of this article using the SQLcl implementation of Liquibase, instead of the Liquibase client.
- That's Not How You Use It!
- Downloads
- Assumptions
- liquibase.properties
- Generating a ChangeLog?
- Create a master.xml
- Applying Changes
- Considerations
Related articles.
- Liquibase : All Articles
- Liquibase : Deploying Oracle Application Express (APEX) Applications
- SQLcl : Automating Your SQL and PL/SQL Deployments using the SQLcl implementation of Liquibase
That's Not How You Use It!
When you look at examples of using Liquibase on the internet they all have a few things in common.
- They are typically examples used to track table changes and not much else.
- Like my examples, they are based on small simple schemas. This always makes sense, but issues arise with some methods when things grow.
- They don't include code objects (procedure, functions, packages, triggers, types etc.).
- If they do include code objects, they assume each version of the code is in a new file. This means you're going to lose the conventional commit history of a file you would normally expect for code. Instead you have to manually diff between separate files.
- They assume people need to rollback changes to previous versions of the database through this mechanism. I think creating a rollback script for each schema change makes sense, but I think it's a bad idea to include it in this mechanism. In my opinion all changes should move forward. So a "rollback" is really a new change applied to the database that reverts the changes. This is especially true as the code related functionality of the Pro edition doesn't support rollback anyway.
- Following on from the previous point, the database code related functionality of the Pro edition doesn't support rollback, so if I have to manually handle rollback, I might as well do the lot.
The major issue for me is the way code objects are managed. This may not affect you if you never have code in the database, but for a PL/SQL developer, this feels like a show-stopper. As a result, I prefer to work using scripts, which are kept in source control, and use Liquibase as the deployment and sequencing mechanism. I'm sure many Liquibase users will not like this, and will think I'm using it incorrectly. That's fine. There's more discussion about script management here.
I'm not hating on Liquibase. This is what feels right to me, but I reserve the right to change my mind!
Downloads
Download the latest version of the software from here. Unzip it into a directory of your choice.
Download the ojdbc8.jar file from here and place it in the "lib" directory of the Liquibase installation you just unzipped.
Assumptions
For the examples here, we have a file system that look like one of these, depending on the OS used. Adjust as required.
Rem Windows c:\software\liquibase\liquibase-3.10.0 : The Liquibase software. c:\git\my_project\changelogs : The location for the individual Liquibase changelogs. c:\git\my_project\scripts : The location of any scripts referenced by Liquibase. c:\git\my_project\changelogs\changelog_master.xml : The master.xml that maintains the changelog order. # Linux /software/liquibase/liquibase-3.10.0 : The Liquibase software. /git/my_project/changelogs : The location for the individual Liquibase changelogs. /git/my_project/scripts : The location of any scripts referenced by Liquibase. /git/my_project/changelogs/changelog_master.xml : The master.xml that maintains the changelog order.
You have Liquibase in your path. Notice I use the UNIX-style separator throughout the rest of the article. Liquibase and Windows don't really care about the separator. A path beginning with "/" is assumed to be relative to the current drive.
Rem Windows set PATH=c:\software\liquibase\liquibase-3.10.0;%PATH% set JAVA_HOME=c:\Program Files\Java\jre1.8.0_221 # Linux export PATH=/software/liquibase/liquibase-3.10.0:$PATHexport JAVA_HOME=/u01/java/latest:$PATH
liquibase.properties
When calling Liquibase you will need to include a number of parameters.
liquibase --driver=oracle.jdbc.OracleDriver ^ --classpath="/software/liquibase/liquibase-3.10.0/lib" ^ --changeLogFile="/git/my_project/changelogs/changelog_00001.xml" ^ --url="jdbc:oracle:thin:@localhost:1521/pdb1" ^ --username=test ^ --password=test ^ {command}
This is a little clumsy, so you can replace all the parameters with a "liquibase.properties" file. If we keep that name and it is present in our current location it will be used, but we will probably need several to cope with different schemas. We create a file called "test_liquibase.properties" with the following contents.
driver: oracle.jdbc.OracleDriver classpath: /software/liquibase/liquibase-3.10.0/lib url: jdbc:oracle:thin:@localhost:1521/pdb1 username: test password: test
We can now run commands using the "--defaultsFile" parameter.
liquibase --defaultsFile="/software/liquibase/test_liquibase.properties" {command}
Generating a Changelog?
If you are starting to use Liquibase against an existing project, you can generate an initial changelog that represents your day 1 state. To demonstrate this, let's create some objects in a clean schema called "TEST".
create table tab1 ( id number, description varchar2(50), constraint tab1_pk primary key (id) ); create sequence tab1_seq; insert into tab1 (id, description) values (tab1_seq.nextval, 'Description for ' || tab1_seq.currval); commit; create or replace function get_tab1_count return number as l_count number; begin select count(*) into l_count from tab1; return l_count; end; /
We generate an initial changelog to capture the state of our initial schema contents.
liquibase --defaultsFile="/software/liquibase/test_liquibase.properties" --changeLogFile="my_changelog.xml" generateChangeLog
The resulting changelog looks like this. Notice it doesn't contain definitions of stored code (procedures, functions, packages, triggers) and the seed data is also missing. It's also worth noting that the object definitions are XML-based, rather than DDL and DML.
<?xml version="1.1" encoding="UTF-8" standalone="no"?> <databaseChangeLog xmlns="http://www.liquibase.org/xml/ns/dbchangelog" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xsi:schemaLocation="http://www.liquibase.org/xml/ns/dbchangelog http://www.liquibase.org/xml/ns/dbchangelog/dbchangelog-3.9.xsd"> <changeSet author="tim (generated)" id="1571293572660-1"> <createTable tableName="TAB1"> <column name="ID" type="NUMBER"> <constraints primaryKey="true" primaryKeyName="TAB1_PK"/> </column> <column name="DESCRIPTION" type="VARCHAR2(50 BYTE)"/> </createTable> </changeSet> <changeSet author="tim (generated)" id="1571293572660-2"> <createSequence maxValue="9999999999999999999999999999" sequenceName="TAB1_SEQ" startValue="21"/> </changeSet> </databaseChangeLog>
For this reason I don't believe generating an initial changelog makes sense for all but the simplest projects. Instead, have a fixed point in time and start using Liquibase from there onward.
Create a master.xml
The master.xml, also known as a controlfile or master index, is an ordered list of changelogs. When we start using Liquibase we can just create an empty one manually with the following contents.
<?xml version="1.0" encoding="UTF-8"?> <databaseChangeLog xmlns="http://www.liquibase.org/xml/ns/dbchangelog" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xsi:schemaLocation="http://www.liquibase.org/xml/ns/dbchangelog http://www.liquibase.org/xml/ns/dbchangelog/dbchangelog-3.9.xsd"> </databaseChangeLog>
Every changelog that happens from now on has to be referenced in order from this master.xml.
We can create one that references every changelog in a specific directory, but this requires the changelogs to be named in a way that the application order is preserved, which could prove problematic. It feels safer to avoid this approach unless you are extremely disciplined.
<?xml version="1.0" encoding="UTF-8"?> <databaseChangeLog xmlns="http://www.liquibase.org/xml/ns/dbchangelog" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xsi:schemaLocation="http://www.liquibase.org/xml/ns/dbchangelog http://www.liquibase.org/xml/ns/dbchangelog/dbchangelog-3.9.xsd"> <includeAll path="/git/my_project/changelogs"/> </databaseChangeLog>
Applying Changes
We have a script called "tab1_seq.sql" in our "/git/my_project/scripts" directory with the following contents.
-- tab1_seq.sql create sequence tab1_seq;
The comment line at the start of each SQL file is the file name. This isn't necessary, but it makes things a clearer when looking at the changelog.
We have a script called "tab1.sql" in our "/git/my_project/scripts" directory with the following contents.
-- tab1.sql create table tab1 ( id number, description varchar2(50), constraint tab1_pk primary key (id) ); insert into tab1 (id, description) values (tab1_seq.nextval, 'Description for ' || tab1_seq.currval); commit;
We have a script called "get_tab1_count.sql" in our "/git/my_project/scripts" directory with the following contents.
-- get_tab1_count.sql create or replace function get_tab1_count returns number as l_count number; begin select count(*) into l_count from tab1; return l_count; end; /
If we want to apply these as a change to the database, so we create a changelog for them. Let's call this "changelog_00001_setup_tab1.xml" and place it in the "/git/my_project/changelogs" directory.
<?xml version="1.0" encoding="UTF-8"?> <databaseChangeLog xmlns="http://www.liquibase.org/xml/ns/dbchangelog" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xsi:schemaLocation="http://www.liquibase.org/xml/ns/dbchangelog http://www.liquibase.org/xml/ns/dbchangelog/dbchangelog-3.9.xsd"> <changeSet author="tim" id="01_tab1_seq"> <sqlFile dbms="oracle" endDelimiter=";" path="../scripts/tab1_seq.sql" relativeToChangelogFile="true" splitStatements="true" stripComments="false"/> </changeSet> <changeSet author="tim" id="02_tab1"> <sqlFile dbms="oracle" endDelimiter=";" path="../scripts/tab1.sql" relativeToChangelogFile="true" splitStatements="true" stripComments="false"/> </changeSet> <changeSet author="tim" id="03_get_tab1_count" runOnChange="true"> <sqlFile dbms="oracle" endDelimiter=";" path="../scripts/get_tab1_count.sql" relativeToChangelogFile="true" splitStatements="false" stripComments="false"/> </changeSet> </databaseChangeLog>
There are some things to consider here.
- The entries in the changelog need to be in the correct order. The creation of the seed data requires the sequence, so the sequence creation must come first in the changelog. The function depends on the table, so the table creation has to happen before the function change.
- We need to make sure "splitStatements" is set to false for database code (triggers, procedures, functions, packages, types etc.) or Liquibase will try to split on ";" and errors will be produced when those changes are applied.
- We are using relative paths (relativeToChangelogFile="true"). Explicit paths can be used if required.
- The ID should be unique for all changesets applied to the schema, so you will often see people using a GUID.
We add a reference to this new changelog into the master.xml, so it now looks like this.
<?xml version="1.0" encoding="UTF-8"?> <databaseChangeLog xmlns="http://www.liquibase.org/xml/ns/dbchangelog" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xsi:schemaLocation="http://www.liquibase.org/xml/ns/dbchangelog http://www.liquibase.org/xml/ns/dbchangelog/dbchangelog-3.9.xsd"> <include file="./changelog_00001_setup_tab1.xml" relativeToChangelogFile="true"/> </databaseChangeLog>
We can then apply the change using the following command. This is run on Windows, hence the "^" for the line continuation. For Linux replace this by "\".
liquibase --defaultsFile="/software/liquibase/test_liquibase.properties" ^ --changelog-file="/git/my_project/changelogs/changelog_master.xml" ^ update
This will create the objects associated with our change, and create some Liquibase objects to track the changes that have been applied.
column object_name format a30 select object_name, object_type from user_objects order by 1, 2; OBJECT_NAME OBJECT_TYPE ------------------------------ ----------------------- DATABASECHANGELOG TABLE DATABASECHANGELOGLOCK TABLE GET_TAB1_COUNT FUNCTION PK_DATABASECHANGELOGLOCK INDEX TAB1 TABLE TAB1_PK INDEX TAB1_SEQ SEQUENCE 7 rows selected. SQL>
If we describe the DATABASECHANGELOG
table we can see the information that's captured.
SQL> desc databasechangelog Name Null? Type ------------- -------- ------------- ID NOT NULL VARCHAR2(255) AUTHOR NOT NULL VARCHAR2(255) FILENAME NOT NULL VARCHAR2(255) DATEEXECUTED NOT NULL TIMESTAMP(6) ORDEREXECUTED NOT NULL NUMBER(38) EXECTYPE NOT NULL VARCHAR2(10) MD5SUM VARCHAR2(35) DESCRIPTION VARCHAR2(255) COMMENTS VARCHAR2(255) TAG VARCHAR2(255) LIQUIBASE VARCHAR2(20) CONTEXTS VARCHAR2(255) LABELS VARCHAR2(255) DEPLOYMENT_ID VARCHAR2(10) SQL>
We display just the ID
and FILENAME
from the table. We can see the three changesets from our changelog.
column id format a17 column filename format a60 select id, filename from databasechangelog; ID FILENAME ----------------- ------------------------------------------------------------ 01_tab1_seq /git/my_project/changelogs/changelog_00001_setup_tab1.xml 02_tab1 /git/my_project/changelogs/changelog_00001_setup_tab1.xml 03_get_tab1_count /git/my_project/changelogs/changelog_00001_setup_tab1.xml SQL>
We then get a second set of changes.
We have a script called "tab2_seq.sql" in our "/git/my_project/scripts" directory with the following contents.
-- tab2_seq.sql create sequence tab2_seq;
We have a script called "tab2.sql" in our "/git/my_project/scripts" directory with the following contents.
-- tab2.sql create table tab2 ( id number, description varchar2(50), constraint tab2_pk primary key (id) );
We create a new changelog for this called "changelog_00002_setup_tab2.xml".
<?xml version="1.0" encoding="UTF-8"?> <databaseChangeLog xmlns="http://www.liquibase.org/xml/ns/dbchangelog" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xsi:schemaLocation="http://www.liquibase.org/xml/ns/dbchangelog http://www.liquibase.org/xml/ns/dbchangelog/dbchangelog-3.9.xsd"> <changeSet author="tim" id="01_tab2_seq"> <sqlFile dbms="oracle" endDelimiter=";" path="../scripts/tab2_seq.sql" relativeToChangelogFile="true" splitStatements="true" stripComments="false"/> </changeSet> <changeSet author="tim" id="02_tab2"> <sqlFile dbms="oracle" endDelimiter=";" path="../scripts/tab2.sql" relativeToChangelogFile="true" splitStatements="true" stripComments="false"/> </changeSet> </databaseChangeLog>
We add a reference to this new changelog into the master.xml, so it now looks like this.
<?xml version="1.0" encoding="UTF-8"?> <databaseChangeLog xmlns="http://www.liquibase.org/xml/ns/dbchangelog" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xsi:schemaLocation="http://www.liquibase.org/xml/ns/dbchangelog http://www.liquibase.org/xml/ns/dbchangelog/dbchangelog-3.9.xsd"> <include file="./changelog_00001_setup_tab1.xml" relativeToChangelogFile="true"/> <include file="./changelog_00002_setup_tab2.xml" relativeToChangelogFile="true"/> </databaseChangeLog>
We can then apply the change the same way we did the first time.
liquibase --defaultsFile="/software/liquibase/test-liquibase.properties" ^ --changelog-file="/git/my_project/changelogs/changelog_master.xml" ^ update
Liquibase knows it applied the first changelog already, so it applies only the second. We can see the new objects present in the database.
column object_name format a30 select object_name, object_type from user_objects order by 1, 2; OBJECT_NAME OBJECT_TYPE ------------------------------ ----------------------- DATABASECHANGELOG TABLE DATABASECHANGELOGLOCK TABLE GET_TAB1_COUNT FUNCTION PK_DATABASECHANGELOGLOCK INDEX TAB1 TABLE TAB1_PK INDEX TAB1_SEQ SEQUENCE TAB2 TABLE TAB2_PK INDEX TAB2_SEQ SEQUENCE 10 rows selected. SQL>
We query the DATABASECHANGELOG
table, and we can see the two changesets from our second changelog have been applied.
column id format a17 column filename format a60 select id, filename from databasechangelog; ID FILENAME ----------------- ------------------------------------------------------------ 01_tab1_seq /git/my_project/changelogs/changelog_00001_setup_tab1.xml 02_tab1 /git/my_project/changelogs/changelog_00001_setup_tab1.xml 03_get_tab1_count /git/my_project/changelogs/changelog_00001_setup_tab1.xml 01_tab2_seq /git/my_project/changelogs/changelog_00002_setup_tab2.xml 02_tab2 /git/my_project/changelogs/changelog_00002_setup_tab2.xml SQL>
Considerations
- As mentioned previously, I don't think you should generate a changelog from an existing schema. I think you should only track changes from the day you start using Liquibase onward.
- All change scripts should have an equivalent "undo" script, but I don't like the concept of rollback as described in Liquibase changesets. Instead I feel a better solution is "fix forward". If a change has a problem and needs to be backed out, the "undo" script should be included as a new change in a changelog, and applied via the master.xml, so any form of rollback/undo is still a move forward. This includes reverting code object versions.
- Everything should be in version control, which is why I've included "git" in the paths. The changelogs and the code don't have to be in the same repository though, depending on your requirements.
- Management of the changelog is important, because it is the serialisation of changes to the database. How you manage the changelog is an interesting process.
- This requires some discipline. So should your manual process, but I suspect many people are more lax than you can be when using an automation tool like this.
- I'm purposely keeping this scripts-based. I know a lot of people will not like this and will prefer to use the XML tag approach to creating changelogs. It just doesn't work for me at this point. It also doesn't support a whole bunch of object types I care about.
- Deciding how to approach different schema and privileged accounts takes some thought. In some cases you will need to coordinate changes to multiple schemas, including some via privileged accounts.
- I don't like the thought of relying on file names to control the apply order, but I know some other products encourage that. This is a personal choice, and the example changelogs here are named with an order in mind, even though we are not using it.
- Every change needs to be in a separate changeset, or this can affect the way failures are handled. That may feel odd to people who like to combine multiple changes into a single build scripts. This is one concession I'm prepared to make.
- There is more functionality in the Pro edition, but it still feels like it suffers from the same issues to me. The added metadata is fine, but the source control issue is still present, and it doesn't include any meaningful rollback functionality.
For more information see:
- Liquibase
- Liquibase Best Practices
- Liquibase : All Articles
- Liquibase : Deploying Oracle Application Express (APEX) Applications
- SQLcl : Automating Your SQL and PL/SQL Deployments using the SQLcl implementation of Liquibase
Hope this helps. Regards Tim...