Folks,
I have a situation where I created a rails migration file which updates the value of an attribute in a model. I ran this migration on several production deployments and then realized the table was never updated due to a typo. What is the best practice to fix this, should I be updating my current migration file with the fix and commit it and then re-run the migration OR should I create a new migration file and remove this migration.
If you ran a migration and it succeeded and then you update your current migration file and try to re-run the migrations, Rails won't update your schema because the migration version will be the same as the current db version.
If you do a rake db:migrate:status you should be able to see the list of migrations and the ones that have been applied or not.
Based on that, if your migration says status: down it means that it still hasn't been applied to the database.
If it has already been applied to the db you can have a number of options.
rake db:rollback update your file then re-run that migration
Create a new file that fixes the typo. Note that even removing the migration with the wrong data won't fix your database depending on what you did to the table and how you fixed it on the new migration.
Related
I feel like this must be a question that has been answered in the past but I could not find anything with Google.
According to the Rails guides:
There is no need (and it is error prone) to deploy a new instance of an app by replaying the entire migration history. It is much simpler and faster to just load into the database a description of the current schema.
However, when you load the database via the schema file there would be no data in the schema_migrations table and therefore any migrations that are added to the project in the future cannot be run without first running the entire migration history.
Am I missing something here? If I create a new instance of the Rails DB from a schema file how can I run any future migrations against it?
It is preferred to set up a new database with a schema load rather than running old migrations, and it should be expected that the schema_migrations table is updated correctly when running a rake db:schema:load. When you run rake db:schema:load the schema_migrations table should be populated based on the value of the version key-pair passed to the ::define method in the db/schema.rb file.
For example:
ActiveRecord::Schema.define(version: 20161208214643) do
end
should update the schema_migrations table up to the migration listed in version. The version value should be populated in the db/schema.rb file automatically after running a rake db:schema:dump or rake db:migrate (which invokes db:schema:dump). This allows for any future migrations to be run normally, as should be expected.
The inserts into the schema_migrations table are handled by the assume_migrated_upto_version method, as seen in the following files:
https://github.com/rails/rails/blob/5-0-stable/activerecord/lib/active_record/schema.rb#L52
https://github.com/rails/rails/blob/5-0-stable/activerecord/lib/active_record/connection_adapters/abstract/schema_statements.rb#L1021
2 months ago, one of my migration files has been removed. The thing is when I rake db:migrate my database in localhost the migration is missing. I can Not create a new migration because the production database was migrated with the missing file. So everytime I import the production database in local and I run a migration it says:
Field already exists....
I saw some comment which says I should modify the schema_migrations table with the corresponding version of the migration file.
Do you have any deeper explanation about this issue.
Thanks
the schema_migrations table holds all the version numbers of all the migrations that have been run, or that the system thinks have been run at least. When you do rake db:migrate, rails looks at all the files in db/migrate, and ignores the ones whose numerical prefix is in schema_migrations.
So, if you have a migration that has actually already been run, but the system thinks it hasn't, you just need to add its version number into schema_migrations.
Open the relevant database console, and type
insert into schema_migrations (version) values("<the number from the start of the migration file>");
obviously, replace the part in < > with the proper number.
Then, db:migrate will no longer try and run that migration.
I'm trying to make a new rails project with an existing database.
I create the projects, configure the database.yml and can successfully do a db:schema:dump
This is all based on what I read on this site
http://blog.joelberghoff.com/2013/02/06/ruby-on-rails-tutorial-creating-a-rails-instance-from-an-existing-mysql-db/
And a few others.
After I do the dump, I copied the schema.rb into db/migrate and renamed to 001_createdatabase.rb as indicated in his tutorial.
Everything seems fine. However I cannot run the site as it tells me i have migrations pending. and the db:migrate:status indicates its that 001_createdatabase.rb that I created.
So based on that:
Was creating that 001_createdatabase migration from the schema.rb the right thing to do?
If it was how do I get rails to understand that it doesn't need to run that one as its already done?
Where does rails check to see if it has pending migrations. Obviously it must check the db/migrate folder and compare it to what?
Thanks
It would seem that in order to get credit for having run the migration, you need to actually run the migration. The migration, 001_createdatabase.rb, represents the migration that will take you from having an empty database -- no tables, no data -- to having your first version of the database. When you run rake db:migrate, rails checks to see if every file in the db/migrate directory has been run by looking for the numerical part of the migration file name in the schema_migrations table.
In your case, you've not run the migration, there is no entry in the schema_migrations table, so rails thinks, rightly, there is a migration to run.
To fix the issue, you can put the record that the migration has been run into schema_migrations or actually run the migration. You may need to create the schema_migrations table, rails will create it as needed.
If you want to run the migration, either drop your database (or create a new one -- you can drop the old one when this works), and point your rails application to the new database. run rake db:create and rake db:migrate. When it completes, you should have a new copy of your database, but one that has been created by the rails migration process. The schema_migrations table should have a record with 1 in the version column.
I created a migration and ran it in a git branch but then did not merge the branch. However, the db columns from that migration continue to exist and I in fact still need them. Since there's no migration file for them (because I didn't merge the branch), I created a new migration but obviously can't run it, because the columns already exist. I therefore renamed the file by giving it a number that made it seem as if it was an earlier migration -- I thought Rails would think that the migration had been run if I changed the number on the file. However, that hasn't worked. Now anytime I run a migration, it's getting aborted because I have this migration file in it that hasn't run but for which the columns already exist. I can delete this file but that won't help me when I push to production.
What do you do in this situation?
I'm using postgres if it matters
rake db:rollback
Then generate the migration in your other branch and
rake db:migrate
Rails checks if a migration has already been run by comparing the migration numbers of the files from db/migrate/ directory with the numbers it finds in the schema_migrations table. If you want to pretend that the migration has already been run, just manually insert the migration number into this table.
However, this is just an ugly workaround, it definitely should not be something to use often - and never use it in production.
Does rake db:migrate only add new migrations, or does it drop all migrations/changes and build everything new?
I think rake is throwing an error because it is trying to access a table attribute in migration 040 that was deleted in migration 042. somehow my DB and rake are out of synch and I want to fix them.
for you experts out there - is it common for rake to get out of synch with migrations? how can I avoid this (no, I do not hand-edit my schema or rake files).
When you use rails migrations, a table called schema_migrations is automatically created, which keeps track of what migrations have been applied, by storing the version number of each migration (this is the number that prefaces the migration name in the file name, ie db/migrate/_20090617111204__migration.rb). When you run rake db:migrate to migrate up, only migrations which have not been run previously (ie. their version is not contained in the table) will be run (for this reason, changing a migration that's already been executed will have no effect when running db:migrate). When migrating down, all versions found in schema_migrations that are greater than the version you are rolling back to will be undone.
Everytime you create a migration using scripts (like script/generate model ...) a new migration is added to the correct directory ready to be synched with the real database.
Actually rake db:migrate just checks which missing migrations still need to be applied to the database without caring about the previouse ones.
Of course if you modify the database using other ways is common to obtain out-of-synch things because as you said you can find yourself applying a migration to something that is changed underneath.
A migration means that you move from the current version to a newer version (as is said in the first answer). Using rake db:migrate you can apply any new changes to your schema. But if you want to rollback to a previous migration you can use rake db:rollback to nullify your new changes if they are incorrectly defined. Caution: by doing so your data will be lost.