db-maintain has the notion of repeatable scripts, such as stored procedures/triggers. When there is a change to such a script it needs to be rolled out again. Flyway seemingly always checks for the filename convention. So to have a repeatable script we might need to rename the file. Is there a more efficient way of doing this?
Flyway started to support repeatable migrations in version 4.0. Repeatable migrations are reapplied every time their checksum changes and can be maintained as single files in version control.
It is your responsibility to ensure the same repeatable migration can
be applied multiple times. This usually involves making use of CREATE
OR REPLACE clauses in your DDL statements.
More details here.
This issue has come up in the Issue Tracker and in this other question.
There is currently no out of the box support for this.
Personally I would
package the repeatable actions in a stored procedure or trigger and add it to the DB as part of a regular migration
make sure this procedure/trigger gets invoked once per migration after that (could be as little a one-line statement)
ensure changes to the procedure/trigger necessary after that also happen as part of regular migrations
If that doesn't do it, feel free to star the Issue and comment with details about your use case.
Update: Repeatable scripts are now fully supported as of Flyway 4.0. See https://flywaydb.org/documentation/migration/repeatable
Related
I use flyway 8.5.0 and I want that my beforeMigrate or my afterMigrate sql is reported in the history table. Is this feaseable? or is there any config to setup this?
Then an other question: my repetable only runs when they change (checksum) but for my understanding the repetible sql should run every time. not so?
The beforeMigrate and afterMigrate SQL wont appear in your history table. If you look at the tutorial example for callbacks you can see that beforeMigrate can be called before the schema history table is created which would cause issues if it was trying to add itself to it. Additionally, I'm assuming these will be mostly static executions and would not really be part of the version history.
https://flywaydb.org/documentation/tutorials/callbacks
For repeatable, no they are only applied when the checksum has changed.
Repeatable migrations are very useful for managing database objects whose definition can then simply be maintained in a single file in version control. Instead of being run just once, they are (re-)applied every time their checksum changes.
https://flywaydb.org/documentation/tutorials/repeatable
I'm investigating how Flyaway works to see if it's suitable for our upgrade and migration needs.
Can somebody tell me if Flyaway automatically handles precedence so that for example, if a table has a trigger, it creates the table before attempting to create the trigger, or is the user responsible to manually creating and ordering the actual SQL Statements required in each schema update?
The latter - Flyway does not keep knowledge of the dependencies between database objects (as it's a very large and gnarly problem, and it's highly db-dependent). Which database do you use? A tool like Redgate Schema Compare will generate the script with all the correct dependency orderings for you.
The normal way of dealing with Doctrine Migrations is via the standard Commands - during development one runs the commands manually to e.g. run diffs and apply the migrations, and deployment typically involves applying them the by the same approach but automatically. Occasionally when working in a team on a local instance there are new migrations, but I've updated my source from version control rather than done a deployment, so I need to apply the new migrations manually, and I need to know that I need to do that! An improvement could be to display a warning on a rendered webpage that migrations are out of sync and action needs to be taken.
Is there a way to access the Migrations API directly in PHP/Symfony code, so that I could detect a mismatch between committed and applied migrations? I haven't found any documentation about that. I've had an initial poke around the code and it seems heavily skewed towards Commands (reasonably enough).
Firstly, updating your source code from version control, is a deployment too, and applying Doctrine Migrations should be part of that. You should create a check list of all the steps you need to do during a deployment, including rollbacks. Depending on the complexity of the application, many things could go wrong.
To answer you question, you can execute, in your code, a diff migration with the Process component and parse the output to determine if there're migrations to be applied.
I am currently using the Flyway Command Line tool to manage our scripts which will be invoked via our release process triggered from our CI Build server.
The issue is I have 274 repeatable migrations covering package specs, package bodies, functions, procedures, views and materialised views.
When I run migrate everything works as expected with migrations executing followed by any changed repeatable migrations but lets say in the next release we want to delete a object which one of the repeatable migrations maintains. For example, we want to remove the repeatable script that defined ProcedureOne (ie R__ProcedureOne.sql).
To do this I would have a new migration script (V3.1.5.1.01__DropProcedureOne.sql) but I would also remove the repeatable migration script so the object isn't created again and maintained.
However, executing flyway info shows the R__ProcedureOne.sql script with the status of MISSING.
While I agree it is missing it is a deliberate action to have it deleted as it is no longer required opposed to being misplaced.
I am aware of the migrate option ignoreMissingMigrations but i think using this introduces risk and could mask genuine missing files.
What is the general guidance on how best to remove a repeatable scripts?
I suggest that you simply keep the file but make it empty (ie zero bytes). Alternatively have a comment in the file that explains that the object it represents has been removed.
As for actually removing it, another option from what you suggested could be to update the repeatable migration to remove itself then update again to zero length. This has the advantage of being able to be replayed into an empty database; since repeatable migrations are applied after versioned the procedure in your example won't exist to be dropped. The disadvantage is running two migrations.
I'm wondering if Flyway allows the use a master file to re-enforce the execution orders of procedures, functions, views and triggers in repeatable migrations? Currently I have an in-house tool that generates master files for these objects so that they can be installed without errors (resolves the inter-dependency between them).
Has anyone had similar experience and how did you handle this?
Thanks!
When Flyway finds multiple repeatable migrations that need to run (because they are new or have been modified), Flyway executes them in the alphabetical order of their descriptions.
This should easily allow you to achieve something similar to what your in-house tool did with its master file.