CODEWITHSUNDEEP
ruby-on-railsoctopus
Kim
Kim

Reputation: 2176

rails - database sharding

I have the following piece of code which copies data from an image table in database A (shard_A) to another table in database B (shard_B)

Here is the structure of the image table in both databases

Image table (shard_A)   -   Image table (shard_B)
- image_id              -   image_id 
- submission_fk         -   sub_fk 
- image_desc            -   img_desc

The column names of the 2 tables are slightly different

  @test_shard_A = Image.using(:shard_A).find(:first, :conditions => ["submission_fk =?", @sub.id ])

  Octopus.using(:shard_B) do

    Image.create(
    :image_id => @test_shard_A.image_id,
    :sub_fk => @test_shard_A.submission_fk,
    :img_desc => @test_shard_A.image_desc
    ) do |primary|
    primary.img_id = @test_shard_A.image_id
    end

  end

The above is giving me the following error message:

unknown attribute: sub_fk

What am I doing wrong?

Upvotes: 0

Views: 560

Answers (1)

Dutow
Dutow

Reputation: 5668

ActiveRecord loads metadata (table information) when the first query is executed for a table. When you start your application with a query to shard_A, it checks the table structure, and constructs your Image class according to it, with submission_fk.

When you change shards, it won't check the table structure again, so it assumes that there is a submission_fk, and assumes no sub_fk, since it doesn't know about that.

I suggest you to use the same structure for all shards, but if you can't do that, you may force ActiveRecord to reload the table information - I never used Octopus, but I think it will work:

Image.reset_column_information

Note that it will degrade the performance of the application, since reloading the metadata is a slow operation.

Upvotes: 2

Related Questions