Helps
ActiveRecord save race condition in Resque
We have a Rails app that is saving a new ActiveRecord object, and then immediately afterwards, enqueueing a Resque job that looks up that record by its ID. Sometimes, the lookup inside the resque job fails – it claims no such row with that ID exists.
If we put the lookup into a loop that catches that exception and retries the exact same lookup, it always eventually succeeds, usually after 1-2 seconds.
The ID exists on the ActiveRecord object before the job is enqueued, which (I think) means that the save transaction has completed. But the resque job is using a different database connection, so maybe there's some weird caching going on? But here Google has failed us.
Any ideas? We could leave it in a loop that catches the exception, sleeps 1 second, and retries, but: ugh.
Some ideas from the audience:
There is a connection object that can be queried about transaction states.
Wait for transaction count increment/decrement?
Could be a problem with the way that sqlite fakes nested transactions.
Best practices for individual users on GitHub
Usually we have a single GitHub account for a project that contains a key for each of the machines being used on a project. On our project we want to use a separate account for each person on the project. Is there a way to do this without a lot of trouble shuffling SSH keys around?
Pivotal Git Scripts may have some tools for this.
You can also use HTTPS URLs and enter username/password on each commit.
Interestings
Lobot – Now with per-project chef recipes
If you need to write your own chef recipes to install your project's dependencies, you can add a cookbooks directory to the root of your project. Make sure to delete the cookbook_paths section from your lobot.yml (to use the default values), or add ./chef/project-cookbooks to the cookbook_paths section.
So, to have a bacon recipe, you should have cookbooks/pork/recipes/bacon.rb file in your repository.
Stately – A font of the US states
http://intridea.github.com/stately/
For all those times you need a vector representation of the USA and to color the states separately.
Apiary – Public API documentation
Takes a well formed file and generates really usable REST API documentation. Integrates with Github for collaboration
Put the Resque enqueue into an after_commit hook and not an after_save hook.
February 7, 2013 at 4:36 pm
I second what JGeiger said. If you enqueue the job from after_save, the transaction isn’t committed, and thus isn’t visible to other processes. But after_commit happens after the commit, so the ID is visible.
February 8, 2013 at 1:11 am