Editing Training 2022 Linux team/ (section)

==Take backup on Simple storage (S3)  23 Aug 2025 - Tetra Support Staff - Tushar==

{{#ev:youtube|8wLxa2OKINE|640}}

'''Video summary:''' "Take backup on Simple storage (S3)" -- a hands-on walkthrough of backing up server data (specifically database dumps) to S3-compatible object storage using rclone. Covers the AWS-side prerequisites: creating a dedicated IAM user in AWS IAM with full S3 access, selecting the "third-party tool access" user type (rather than AWS Management Console access) when creating it, and generating that user's access key ID and secret access key for use by rclone. Walks through installing rclone via a curl-downloaded install script, then configuring a named "remote" (visible later via `rclone config show`) pointing at the S3-compatible endpoint -- demonstrated using DigitalOcean-style object storage as one example endpoint, but explicitly generalized: rclone/most S3 clients (including the S3 Browser GUI tool also demoed) support any S3-compatible provider, not just AWS S3 itself, selectable via an account-type dropdown ("Amazon S3" vs "S3 Compatible Storage"), with the connection URL/endpoint and region needing to match the chosen provider. Explains the actual backup command structure: `rclone sync` (not `copy` -- sync mirrors changes to the destination rather than accumulating duplicate copies each run) with the source being a local database dump file/variable and the destination written as `remotename:bucketname/folderpath`; notes the target bucket and folder are auto-created on first run if the third-party access permissions are set correctly, and that bucket names must be globally unique across all of S3 (not just within the account) -- if the sync command doesn't auto-create the bucket, that's a signal the chosen name is already taken elsewhere and needs to change. Covers two important tuning flags used in Tetra's actual backup scripts: retention logic that deletes all but the 7 most recent incremental backups as new ones land (rolling 7-copy retention), and `--tpslimit 5` to rate-limit the client's own outbound API requests (GET/HEAD checks against the bucket) so as not to exceed the storage provider's request-rate limits, which otherwise produces rate-limit errors during sync -- particularly relevant since `sync` (versus plain `copy`) does more existence-checking against the destination and is more prone to hitting this limit. Also briefly demonstrates the equivalent GUI workflow in S3 Browser for managing multiple S3-compatible accounts/credentials side by side.