ES initial index rake task requires too much babysitting
@jnijhof is spending a lot of time the past month trying to get ES indexed for gitlab.com. This is holding up deploys, which is not good.
To make the rake task manageable @jnijhof wrote a meta-script to divide the work. Shouldn't this be the job of the rake task?
Indexing 1,000 repos in a row lets the rake process grow to 5GB. The more repos get indexed in a row, the bigger the rake process grows. This is not OK. (Suggested fix for rake task memory grow: #296 (closed).)
We should never ask an admin of one of our customers to spend this much of their time to enable just one GitLab feature.