Reduce memory need of the app in staging/prod

[yrodiere]

The cluster we deploy to is rather resource constrained, in particular when it comes to memory. Let's try to improve where we can.

We can't do much for Elasticsearch, which we're frankly starving at this point. 2GB is not a lot for Elasticsearch.

The app itself is running in JVM mode. Going native could reduce the needs significantly; but last I checked, it didn't work for JGit. Let's try again, and report any problem?

Alternatively, we could change how indexing happens in prod/staging. We currently reindex in the "main" container, which means it needs to request 1/2GB of memory constantly. If we used an OpenShift "Cron" instead, we could have the "main" app do search only, so that it doesn't require much memory and thus wouldn't request as much memory, while the "Cron" container could request as much as it needs.

[yrodiere]

The app itself is running in JVM mode. Going native could reduce the needs significantly; but last I checked, it didn't work for JGit. Let's try again, and report any problem?

I'd go for this first. It seems much simpler to implement/maintain.