Skip to content

[SPIKE] Determine Backup Conventions #5428

@jperson1

Description

@jperson1

Backup Conventions

As a general reminder, please update any docs after decisions are made.

A Little Context

We backup everything every two hours during work hours (7am to 7pm or so). The files are stored in directories named month-day-hour, so they end up overwriting each other after a year has passed. Thus, the bucket grows only a little bit in size once the whole year has been stored. That is a downright silly amount of data.

Retention

How long should we realistically expect to need a backup? If an issue is found we'll use the most recent "good" backup, which may be as little as a few hours old. Our backups lose value rapidly, since during a restore users will lose all progress made since the backup. Going back more than... A week? Is detrimental.

Perhaps we keep all backups in the last week, one a day for the month, and none before that. But storage is cheap, so perhaps we could/should keep a full month.

Deletion

How often should we delete backups? Perhaps on its own action every weekend. Perhaps as an addition to the regular backups.

Naming

Backups folders are named month-day-hour. If we are constantly creating and deleting backups properly, we can instead timestamp them. YearMonthDay-hour, I'd suggest.

Requirements to Consider

Does our ATO say anything about backup cadence or retention? Is there an ADR that discusses? It would be good to do a little discovery as part of these conversations.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    Status

    Backlog

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions