## Proposal Neither our readiness document, opstrace/docs, or runbooks properly document the autoscaling approach and provisioning expectations for the initial deployment. We should write this down somewhere ## Background The following discussion from gitlab-com/gl-infra/readiness!112 should be addressed: - [ ] @mwasilewski-gitlab started a [discussion](https://gitlab.com/gitlab-com/gl-infra/readiness/-/merge_requests/112#note_1018134896): (+1 comment) > What's the plan for the opstrace insfrastructure itself? Is there going to be auto-scaling enabled on any of the components? If not, is there going to be significant over-provisioning? Are there going to be any queues? Are they capped? What other backpressure mechanisms do we intend to put in place? Do we rely on users to retry uploads? Can we explicitly state that users cannot be expected to keep data for longer than a few minutes? > > I'm just thinking out loud. I realize that answers to a lot of these questions are still unknown. As per my comment above, that is totally fine and we can merge the MR without finding answers to them and update the document later.