I recently deployed a website to heroku and I am having problems with the database. I am using ruby on rails and when I first created my local databases I used SQLite. When deploying to heroku I had to change my databases to Postgres and in the process several migration files got removed and changed into a single migration. Now, when I run the postgres database on my localhost, I am still seeing all of the columns in my database, but when I visit the page on heroku one of the columns is missing.
I have checked both my local console and the heroku console and the databases are now different. The local console includes all of the columns in the database and the heroku console is missing one of the columns in the database, but is generating all of the other columns correctly. I have tried running a rake task on heroku and have pushed the most recent changes to heroku.
I tried to to add an additional migration to add the missing column to the database, but whenever I try to rake the migration it tells me that attribute already exists. Any help I can get is appreciated!
The best way to set up your database on heroku is with
rake db:schema:load
From the guides on migrations
Migrations, mighty as they may be, are not the authoritative source
for your database schema. That role falls to either db/schema.rb or an
SQL file which Active Record generates by examining the database. They
are not designed to be edited, they just represent the current state
of the database.
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.
Related
I have created an application with Ruby and Rails. The thing is that when I was develpoing it, I had some problems with the migrations, because I created them but with a wrong syntax. What happened is that I deleted some of the files because sold migrations that didn´t work had the same name than the new ones, but in the middle of that I accidentally deleted some of the migrations (obviously after running rails db:migrate) that the project uses actually. So for instance, i have the Service table, which is related to the Reservation table because Service has reservation_id, but i don´t have the migration file that says AddReservationIdToService.
So now I want to use Heroku for production. the thing is that O have to change to postgresql because Heroku doesn't support sqlite. So i have to run the de:migrate again to create the tables and relationships in the new DB, but I need the files that I explained that I deleted. The question is:
Can I create the migrations manually, so when i run db:migrate for postgres the full structure of the database is created without lacking relations?
You don't really need the migrations to recreate the existing DB -- in fact it's not a good idea to try for a couple of reasons (including the missing migration file problem you encountered). You can simply run:
bin/rails db:schema:load
to populate a new database from the existing schema. If for some reason you haven't got a db/schema.rb checked under version control you can run:
bin/rails db:schema:dump
against the sqlite version to re-create a fresh schema file from the database.
You can also keep your migrations list tidy by occasionally zapping really old migrations, since all the cumulative changes are captured in the schema file.
Yes, you might create another couple of migration files.
Certify you have now the tables you wish locally with your sqlite. Draw these table in a piece of paper (or where it be the best fr you), then check this official API documentation of Rails.
Delete all migrations made before and create another according to the tables you drew.
The workflow is gonna be like:
1) "I need to create a table called Reservation, where is it shown on the documentation?"
2) "I need a table called Service, where is it shown on the documentation?
3) "I need to add a column with a foreign key to service named reservaton_id, how does this documentation says it?
For all this steps above, create the correspondent migration file as you normally have done.
The main difference here is not to run the migration locally. Instead, push your new version app to your heroku remote branch and there you run the migration, like:
heroku run rails db:migrate
Remember to not run this same migration locally because you already have these tables locally.
The last two advise is:
1) If your migration doesn't go as you expect, don't delete the migration file. Instead, run rails db:rollback and try again.
2) Keep tracking your migration files on the same branch of your version control.
I have a rails app on heroku. I recently updated the database by changing a column name and also adding a new column. Both the tasks were done using migration files. The changes were reflected properly on my local system. Also when I ran rake migration on heroku and restarted the app there was no problem. I can see the changes in the database schema by remotely connecting using pgadmin. But I cannot use the columns that are modified. When I edit my code to use the new columns application says that there are no such columns.
Thanks
At some point in my rails development I started making database changes (e.g. dropping or altering columns/tables) without using rails migrations. So now I get errors when I try to deploy my rails app from scratch.
blaine#blaine-laptop ~/tmp/rbjacolyte $ rake db:migrate
(in /home/blaine/tmp/rbjacolyte)
== AddHashToTrack: migrating =================================================
-- add_column(:tracks, :hash, :string)
rake aborted!
An error has occurred, all later migrations canceled:
Mysql::Error: Table 'jacolyte_dev_tmp.tracks' doesn't exist: ALTER TABLE `tracks` ADD `hash` varchar(255)
(See full trace by running task with --trace)
How can I sync my production and development environments with migrations after I've mucked it up by using raw SQL? I want to deploy my rails application without database errors, and I don't want to start from scratch.
The data in the production and development environments match, but the migrations fail. I want a way to 'start from scratch.'
Could I simply delete all of the migrations that I have, and then just start using migrations from now on?
The shortcut way: manually add an entry to schema_migrations for a timestamp that represents a baseline. You can add migrations after that and as long as they don't make any bad assumptions about the db schema they should be able to run just fine. You won't be able to migrate backwards, but that's not a huge problem.
The bigger problem is that you won't be able to make a DB from scratch, which gets to be a pain longer term.
The fix for that is to delete all your existing migrations and create a new one that creates the existing schema. Manually delete everything from the schema_migrations table and put in an entry for this one new migration. After that, you can create new migrations that build on this new baseline and they should apply just fine. You should be able to bootstrap new databases in the normal fashion.
As long as your direct SQL is contained in Rails migrations, there's no problem with using it. Just make sure you implement both the #up and #down methods and you should be good. We've actually taken to using raw SQL as a best practice to avoid problems when models are changed later on. Something like
Foo.create(:name => 'bar')
seems innocuous, until the User model is modified to have
validates_presence_of :baz
At which point the new migration will run against an existing database, but that earlier migration that created the table and added the dummy entry will fail because User fails validation. Just using
execute("insert into foos (name) values ('bar')")
will work fine as long as the later migrations properly populate any new columns they add.
Maybe you could just get rid of all your current migrations, and use rake db:schema:dump to create a new schema.rb file, and manually edit your production database to reflect the changes you've made so far?
I like Veeti's suggestion, with a modification: rake db:schema:dump, then move that file to your development machine. Flatten your Rails migrations so far (see this SO thread on that), get rid of most of your migrations, and re-work your migrations to work, given your new schema.
Get this working on your dev machine, commit and deploy.
If the existing production data is compatible with the development database schema, then I would:
Dump the production data to a file using a program such as mysqldump
Drop the production database
Recreate the production database
Run the migrations against the production database, specifying VERSION=0
Import the production data from the file created at step one
If the schemas aren't compatible then you might be able to follow this process but you'll have to edit the SQL in the file created in the first step to take account of the schema differences.
I was able to push my Ruby on Rails app with MySQL(local dev) to the Heroku server along with migrating my model with the command heroku rake db:migrate. I have also read the documentation on Database Import/Export. Is that doc referring to pushing actual data from my local dev DB to whichever Heroku's DB? Do I need to modify anything in the file database.yml to make it happen?
I ran the following command:
heroku db:push
and I am getting the error:
Sending data
2 tables, 3 records
!!! Caught Server Exception | ETA: --:--:--
Taps Server Error: PGError ERROR: duplicate key value violates unique constraint
"unique_schema_migrations"
I have 2 tables, one I create for my app and the other schema_migrations. The total number of entries among the 2 tables is 3. I'm also printing the number of entries I have in the table I have created and it's showing 0.
Any ideas what I might be missing or what I am doing wrong?
EDIT:
I figured out the above, Heroku's DB already have schema_migrations the moment I ran migrate.
New question: Does anyone know how I can exclude data from a specific table from being pushed to Heroku DB. The table to exclude in this case will be schema_migrations.
Not so good solution:
I googled around and someone else was having the same issue. He suggested naming the schema_migrations table to zschema_migrations. In this way data from the other tables will be pushed properly until it fails on the last table. It's a pretty bad solution but will do for the time being.
A better solution will be to use an existing Rails command which can reset a specific table from a database. I don't think Rake can do that.
Two possible options:
The heroku gem and the taps gem (which it uses to synchronize databases) are both open-source - you could fork them, alter the taps client API to support excluding tables from a push, then alter the heroku gem to use that new option.
You could write a wrapper script that uses pgdump to backup the schema_migrations table, drops that table, heroku pushes the database, then reloads the table.
This is kind of a guess based on the error you're getting, but push looks like it grabs the schema and the data. I'd try push to an empty database.
I've just deployed a Rails 3 beta app to heroku on their new bamboo server. I can now upload data from my local dev machine to the heroku database by doing:
heroku rake db:fixtures:load test/fixtures/my_model.yml
The data is then properly propagated in the Heroku database. Even though I specified a specific data file, it automatically pushes data from my other yaml files. It probably has something to do with my model relationships.
If your databases are out of sync you can always reset the Heroku database before pushing using
heroku db:reset
To push/pull from specific tables
heroku db:pull --tables logs,tags
http://blog.heroku.com/archives/2010/4/21/supporting_big_data_part_1
I have two machines... a development machine and a production machine. When I first brought my rails app onto the production server, I had no problem. I simply imported schema.rb by running rake db:schema:load RAILS_ENV=production. All was well.
So, then on my development machine, I made some more changes and another migration, and then copy the new application over to the production machine. I then tried to update the database by running rake db:migrate RAILS_ENV=production. I get the following error:
"There is already an object named 'schema_migrations' in the database."
I'm thinking to myself, ya no kidding Rake... you created it! I ran trace on rake and it seems as if rake thinks it's the first time it's ever ran. However, by analyzing my 'schema_migrations' table on my development machine and my production machine you can see that there is a difference of one migration, namely the one that I want to migrate.
I have also tried to explicitly define the version number, but that doesn't work either.
Any ideas on how I can bring my production server up to date?
Update:
Let me start off by saying that I can't just 'drop' the database. It's a production server with a little over 100k records already in it. What happens if a similar problem occurs in the future? Am, I to just drop the table every time a database problem occurs? It might work this time, but it doesn't seem like a practical long term solution to every database problem. I doubt the problem I'm having now is unique to me.
It sounds like the 'schema_info' table and the 'schema_migrations' table are the same. In my setup, I only have 'schema_migrations'. As stated previously, the difference between the 'schema_migrations' table on the production server and the development machine is just one record. That is, the record containing the version number of the change I want to migrate.
From the book I read, 'Simply Rails 2', it states that when first moving to a production server, instead of running rake db:migrate, one should just run rake:db:schema:load.
If it matters, I'm using Rails version 2.1.
This is a guess, I admit: I think that because you first ran db:schema:load instead of db:migrate in your production environment, you got the structure of your db, but not the data that migrate populates into your schema_info table. So now, when you run migrate in the production environment, there is no data in schema_info which is why migrate believes that it hasn't run yet (because it hasn't).
That said... you say that you have looked in the "schema_migrations" table, and that there is a difference of one version from dev to production... I haven't heard of that table, although I'm a few months behind on my rails version. Maybe you could try creating a "schema_info" table in the production environment, with a single "version" column, and add a row with the version that you believe your production environment to be on.
If you get "There is already an object named 'schema_migrations' in the database." error message then I suspect that you are using MS SQLServer as your database? (As this seems like MS SQL Server error message)
If yes then which ActiveRecord database adapter you are using? (What is your database.yml file, what gems have you installed to access MS SQL Server database?)
Currently it seems that Rails does not find schema_migrations table in production schema and therefore tries to create it and this creation fails with database error message. Probably the reason is upper/lower case characters in schema_migrations table name - as far as I understand MS SQL Server identifiers are case sensitive.
Depending on the system used in production, I have seen instances where the below does not work:
rake db:migrate RAILS_ENV=production
But where this one does work:
RAILS_ENV=production rake db:migrate
Quirky, I know, but it's worth trying it to see if it makes a difference.
Regarding your update:
I don't understand what the difference is between your production schema_migrations and the dev version. Is there a record in both tables (there should be just 1 column, "version", right) or is there a single record in the dev DB and zero records in production? If there are zero records in the production table, then do this:
ActiveRecord::Base.connection.execute("INSERT schema_migrations (version) VALUES(#{my version number that production is supposedly on})")
Alternatively, you could try dropping the schema_migrations table totally on production:
ActiveRecord::Base.connection.execute("DROP TABLE schema_migrations")
Then, re-running rake db:migrate RAILS_ENV=production. That will run migrations from starting from version 1 though, which is probably not what you're after.
Alternatively alternatively, you could start an IRB session in your production environment, do either a "require" or "load" (I can never remember which, or if it matters) of the migration file that you want to load, and then call MyMigrationClass.up. You would need to manually set the version number in the schema_migrations table after that, as you would still have the problem going forward, but as a quick-fix type of hack, that would work.
I would just drop the DB, add it again and run rake rb:migrate. Brad is correct that when you ran the schema load, it didn't put any records in the schema_migrations table.
This is more complicated of course if there is data you can't lose on the production server. You could get the rake backup tasks (not sure if that is part of core or not) and then run rake db:backup:write on your production database, and then after you get the migrations up to date on production, run rake db:backup:read.
schema_info is from an old version of Rails. schema_migrations is the new kid on the block. You should be able to remove the schema_info table as it'll no longer be used. You'll probably want to search for any issues associated with this name change.
rake db:schema:load will load the database structure from schema.rb. This file is the current representation of the database structure. It's used when you have an empty schema (database) that needs all the tables and indexes creating. It saves you having to run all the migrations. If you have an existing production database with data in, you don't want to run it. As others have said that would be bad!
I know this post was some time ago, but I stumbled across it and it hasn't really been answered. As it comes up on google, here goes.
When you did a rake db:schema:dump (or when this was done for you by the build scripts) it will have put the definition of the migrations table into the schema.rb. At the end of the script, the process will try to create the table again, however it obviously exists already. Just remove the migrations table from the schema.rb before running rake:schema:load and there will be no error message.
You will need to set the version number in the migrations table to subsequently run migrations. So it is important to know what version your schema.rb relates too, or delete all the old migrations (they're safely in your SCM right?)
rake db:migrate RAILS_ENV=production
Use the db:schema:load task just for the first creation, incremental changes should be migrated.