Scylla Manager Agent

• Reads config from scylla API on start
• Listens on Scylla
• CPU pinning
• Agent conf file etc/scylla-manager-agent/scylla-manager-agent.yaml

Scylla Manager Backup Benefits

• Data selection - single table or entire keyspace
• deduplication - saves network and s3 costs
• Data retention - purge if all goes well, or failover if something goes wrong
• Data throttling - control how fast/slow want to upload,
• start, pause, stop, resume
• retries
• visibility - progress can be read from cli, rest api, prometheus metrics

Backup process

Multiple steps are executed sequentially on all nodes in parallel

• Snapshot - take snapshot of data on each node according to config
• Schema - optional [requires cluster info to be added to manager agent] Uploads CQL schema
• Upload - upload to desired s3 path
• Manifest - metadata about the upload
• Purge - removes oldest backup from storage if retention threshold has been reached

Backup prep

• create a storage location for backup
• grant access control to your storage buckets
  • S3: Iam role
• validate manager access to the backup location
  • scylla-manager-agent check-location --location s3:<path/to/backup/location>
• sctool backup -c <cluster-name> -L <s3-location> --retention <retention-period-number-in-days> -s '2022-04-04T15:16:05Z' [schedule time to run backup] -i 24h/7d [interval]
• Also can use --dry-run param to check if the command is correct and scylla is able to access backup loaction, dc and cluster