Skip to content

Aerospike Backup Service 3.6.1 release notes

For the complete documentation index see: llms.txt

All documentation pages available in markdown.

July 24, 2026  |  Download

New in this version

Aerospike Backup Service (ABS) 3.6.1 is built on backup-go v0.11.0. It expands restore-by-timestamp for cross-region and disaster recovery workflows, adds a compact backup policy option for smaller backup files, improves backup observability, and streamlines performance.

Breaking changes

  • Restore job status response values are now lowercase: running, success, failure, and canceled. The previous values were Running, Done, Failed, and Canceled. Update integrations that compare these response values. [BKRS-127]
  • The default service log level changed from DEBUG to INFO. Set service.logger.level to DEBUG explicitly to retain verbose logging. [BKRS-186]
  • The default min-part-size for Amazon S3, Azure Blob Storage, and Google Cloud Storage (GCS) increased from 5 MiB to 50 MiB. At the default parallel-write value of 8, chunk buffers can use about 400 MiB per active namespace. See min-part-size tuning. [BKRS-187]
  • For finite file-limit values, ABS rejects cloud storage configurations that require 10,000 or more chunks per backup file. A file-limit value of 0 disables this validation. Increase min-part-size or decrease file-limit if validation rejects an existing configuration. [BKRS-187]
  • Backup scan parallel is now enforced per routine instead of per namespace. For routines with multiple namespaces, the total concurrent scans across all namespaces cannot exceed parallel. Verify that backup duration still fits within the scheduled backup window after upgrading. [BKRS-160]
  • When the source cluster defines max-parallel-scans, an explicitly configured parallel value cannot exceed it. The cluster limit is shared across active routines. This validation runs at startup and when the configuration changes. [BKRS-238]
  • The aerospike_backup_service_restore_in_progress metric now counts only active restore jobs. Update queries or alerts that relied on completed jobs remaining in this gauge. [BKRS-127]
  • Startup storage checks now verify read and list access for Amazon S3, GCS, and Azure Blob Storage. Missing read or list permissions prevent startup. Write or delete permission failures produce warnings and can cause backup or cleanup operations to fail later. [BKRS-194]

New features

  • Restore-by-timestamp requests can override source or source-name to read backups from a different storage location than the configured routine. This supports cross-region migration and disaster recovery restores. [BKRS-212]
  • When both source and destination are provided in a restore-by-timestamp request, the routine no longer needs to exist in the ABS configuration file. See Restore using routine name and timestamp. [BKRS-212]
  • Added a compact backup policy option. When enabled, ABS skips Base64 encoding for binary large object (BLOB) types, producing smaller backup files. The supported BLOB types are Bytes, HyperLogLog (HLL), RawMap, and RawList. [BKRS-198]
  • A restore-by-timestamp request with policy.unique set to true leaves existing records unchanged. With the default incremental reordering, ABS adds missing records using the newest backed-up version at the target timestamp. Do not set disable-reordering to true when the newest version is required. [BKRS-215]
  • Added support for Azure Blob Storage authentication with a shared access signature (SAS) token in the storage endpoint URL. The URL contains the credential, so protect it as a secret and use a short-lived token with only the required permissions. [BKRS-248]
  • Added the aerospike_backup_service_restore_events_total Prometheus counter with an outcome label (success, failure, canceled) for completed restore jobs. [BKRS-127]
  • Added the aerospike_backup_service_backup_in_progress Prometheus gauge with routine and type labels for running backups. [BKRS-127]

Improvements

  • Improved Aerospike record reader performance. [BKRS-178]
  • Improved restore decoder performance. [BKRS-182]
  • Multi-namespace routines now schedule scans fairly across namespaces. [BKRS-238]
  • The aerospike_backup_service_backup_progress_pct metric now reports fractional progress without rounding. ABS recalculates the estimated record count every 10 minutes during a backup. [BKRS-185, BKRS-218]
  • Clarified restore-by-timestamp API field descriptions in the OpenAPI specification. [BKRS-104]
  • Backup retries now start in a new timestamped directory, preventing path collisions between attempts. [BKRS-195]
  • Relaxed finished timestamp validation so backups created before ABS 3.4 remain restorable. [BKRS-209]
  • Removed the hardcoded localhost:8080 server URL from the generated OpenAPI specification. [BKRS-242]
  • ABS now rescans affected routine storage after successful backups and changes to backup routines or their storage configuration, refreshing last-backup timestamps.

Bug fixes

  • Fixed a race condition that could allow a restore to start while a backup was starting on the same cluster and namespace. [BKRS-242]
  • Fixed public key infrastructure (PKI) cluster authentication when no password is configured. [BKRS-251]
  • Fixed retry handling for transient errors during large GCS uploads. [BKRS-244]
  • Released scan limiter slots when a backup is canceled so later backups can start. [BKRS-114]

For additional details about fixes, improvements, and maintenance in this release, see the release page on GitHub.