We have just restarted the instance. This caused a few minutes of downtime, but it allows the server to get back on track! We should do it more often.

Some optimizations have been made to cope with RAM saturation.

We reverted the optimization of Sidekiq, we went back to 48 threads for all the processing queues.

Reduced ElasticSearch to 1GB of memory.
And some optimizations in Nginx and Puma.

Happy meow !