Storage Growth

What you can build

• Expire application logs before the bucket bill grows

  A log prefix whose BucketSizeBytes climbs each day gets a proposed Lifecycle rule: transition to S3 Standard-IA after the age you set, expiration later. The proposal lists the objects already past each age.

  With Amazon S3, Amazon CloudWatch

• Resize a database volume before WAL runs out of room

  When pg_database_size and the volume series in Grafana project a full disk inside your horizon, the operator files a proposed Jira issue with the date, the series, and the tablespace option.

  With PostgreSQL, Grafana, Jira

• Move a cold bucket to archive storage

  A bucket with steady NumberOfObjects and no request metrics gets a proposed copy to an archive prefix, or a Lifecycle transition to S3 Glacier Flexible Retrieval, with the retrieval terms restated in the proposal.

  With Amazon S3, Amazon CloudWatch

Requirements

• Amazon S3 buckets with a CloudWatch read grant for the AWS/S3 daily storage metrics, and s3:PutLifecycleConfiguration where rules are to be applied
• A PostgreSQL role allowed to run the size functions on the databases you watch
• A Grafana service account token with datasources:read, and alerting provisioning rights if rules are proposed
• A ceiling per target: the volume size, the bucket budget, or the disk behind the database, since none of these systems reports its own

Authentication

Reads metrics with the CloudWatch, PostgreSQL, and Grafana grants you attach, and writes lifecycle rules or issues only with the S3 and Jira grants you approve.

Limits

• S3 daily storage metrics are best effort. A day's point can arrive late or not at all, so each projection states its point count.
• PostgreSQL relpages is refreshed only by VACUUM, ANALYZE, and some DDL. The size functions are read instead, but bloat reads as growth until vacuumed.
• An S3 Lifecycle rule applies to existing objects as well as new ones. A bucket policy cannot stop it. Each proposal names the objects affected.
• It never runs a resize, a VACUUM, or a delete. A resize is a Jira issue for a person. Lifecycle rules apply only on approval.

Questions and answers

What do I approve?

One proposal per bucket, volume, or database: a Lifecycle rule with its prefix and ages, an archive copy with its destination, a resize issue, or an alert rule. Each shows the size series, the fitted line, and the projected fill date. You edit the ages or the destination before approving.

What record is left?

A receipt per approval: the series and projection at the time, the rule or issue created, its id in S3, Jira, or Grafana, who approved, and how to remove the rule. Projections that missed are kept beside the ones that held.

Will it delete data?

Only through a Lifecycle expiration you approved, and the proposal names the prefix and the object count that expiration covers. It never issues a delete itself, never drops a table, and never changes a storage class outside an approved rule.