"Zero downtime deploys?" (documentation)

[mikermcneil]

Goal

Document how to achieve "Near-zero downtime deploys" (near-zero b/c migrations).

e.g.

## Zero downtime deploys?

lorem ipsum… 1-2 paragraphs

How?

• [ ] downtime due to schema migrations is practically unavoidable and from conversations with users, not a huge deal because scheduled queries queue/buffer undeliverable query results at the osquery agent
• [ ] in general the queueing/buffering behavior makes non-zero downtime deploys less painful
• [ ] that said, it can be important for some teams, so documenting what the community has learned from each other could be helpful (under "Docs > Deployment")

[zwass]

I'd consider Fleet deploys right now to be "low downtime". Typically I think folks are experiencing <5m downtime during upgrades (and not losing data from this period due to osquery's buffering). May be useful to try to get some empirical data on this.

To achieve "zero downtime", we would need to architect migrations in such a way that they are backwards compatible to at least the previous version. This is certainly possible but would introduce additional friction to the backend development process.

[fleet-release]

A cloud of calmness, Zero downtime deploys bring peace. Smooth migrations soar.