Feature Request: Keyspace-wide Backups & Flexible Scheduling in vitess-operator

[bluecrabs007]

Summary

We would like to propose enhancements to the VitessBackupSchedule feature introduced via #553 in vitess-operator to improve operational flexibility, reduce bandwidth spikes, and simplify configuration for large deployments.

This request includes:

1. Cluster/Keyspace-wide backup modes (all shards)
2. Flexible scheduling with optional randomized distribution

---

1️⃣ Feature: Backup All Shards in Keyspace / Cluster

Problem

Currently, backups require specifying shard-level configuration. This becomes operationally heavy for:

• Large keyspaces with many shards
• Multi-keyspace clusters
• Environments where consistent backup policies are desired across all shards

Proposal

Introduce higher-level backup modes:

• BackupAllShardsInKeyspace
• BackupAllShardsInCluster

These modes would:

• Automatically discover shards
• Avoid requiring per-shard configuration
• Potentially execute backups in parallel (or configurable concurrency)
• Rework prior sequential implementation to avoid excessive runtime

Goal

Simplify configuration and reduce operational overhead for large Vitess deployments.

---

2️⃣ Feature: Flexible Backup Scheduling (Fixed or Randomized)

Current State

The schedule is currently cron-based and user-specified.

Problem

For large keyspaces, triggering backups for all shards at the same time can:

• Cause significant upload bandwidth spikes
• Create resource contention
• Increase operational risk

We can allow users to configure frequency (e.g., every 24 hours, every 12 hours), and then we:

• Generate per-shard randomized cron schedules
• Persist the selected cron schedule
• Stagger backups across shards

Proposal

Allow users to specify either:

Option A – Fixed Schedule

schedule: "0 1 * * *"  # Every day at 1am

Option B – Frequency-Based (Randomized)

backupEvery: 24h
randomizePerShard: true

Behavior:

• Operator generates a random cron schedule per shard
• Schedule is persisted
• First backup waits until scheduled time
• Backups are evenly distributed over time