How and when to deal with memory leaks in gitaly-ruby
Currently a gitaly-ruby process uses about 50 MB RAM (RSS) after booting. In gitlab.com production, where we are only sending very light traffic to gitaly-ruby at this time, we see most gitaly-ruby processes using 750-3000 MB RSS.
I think we can say that we have a memory leak. When do we make it a priority to deal with this, and how do we want to deal with this?
See https://prometheus.gitlab.com/graph?g0.range_input=1d&g0.expr=gitaly_supervisor_rss_bytes%7Benvironment%3D%22prd%22%7D&g0.tab=0 for a graph illustrating the problem.
Edited by Jacob Vosmaer