For prod deployments use snapshots instead of rebuilding the db on prod

[robinbb]

After separating Prod Indexer deployments to their own branch (#1756), use the database snapshots to start the Indexer.

This will help avoid the problem where prod goes down because of out-of-memory condition caused by a database rebuild.

[jhult]

Loosely related to Granola-Team/infra#23

Perhaps the same upload mechanism will be used for both.

[Isaac-DeFrain]

What is the status of this?

[jhult]

For branching, this is the idea: https://trunkbaseddevelopment.com/branch-for-release/

Tags may be an interesting possibility:

Late creation of release branches
Some teams release from a tag on the trunk and do not create a branch at that time. That in itself is an alternate practice to this one, “branch for release”.

Those teams wait for a bug that needs fixing for a released, before creating a branch from the release tag (if they are not going to just issue another release from the trunk).

Brad Appleton points out that many do not realize that branches can be created retroactively. That is taken advantage of here in the case of bugs after “release from a tag”, or even changes for point releases.

[jhult]

If we want to use a CNAME such as db-snapshots.minasearch.com, the Linode bucket name will need to be that exact CNAME.

[robinbb]

Blocked on #1781 .