proxmox-backup/pbs-datastore
Christian Ebner 6a880e8a44 datastore: implement garbage collection for s3 backend
Implements the garbage collection for datastores backed by an s3
object store.
Take advantage of the local datastore by placing marker files in the
chunk store during phase 1 of the garbage collection, updating their
atime if already present.
This allows us to avoid making expensive API calls to update object
metadata, which would only be possible via a copy object operation.

The phase 2 is implemented by fetching a list of all the chunks via
the ListObjectsV2 API call, filtered by the chunk folder prefix.
This operation has to be performed in batches of 1000 objects, given
by the APIs response limits.
For each object key, lookup the marker file and decide based on the
marker existence and it's atime if the chunk object needs to be
removed. Deletion happens via the delete objects operation, allowing
to delete multiple chunks by a single request.

This allows to efficiently lookup chunks which are not in use
anymore while being performant and cost effective.

Baseline runtime performance tests:
-----------------------------------

3 garbage collection runs were performed with hot filesystem caches
(by additional GC run before the test runs). The PBS instance was
virtualized, the same virtualized disk using ZFS for all the local
cache stores:

All datastores contained the same encrypted data, with the following
content statistics:
Original data usage: 269.685 GiB
On-Disk usage: 9.018 GiB (3.34%)
On-Disk chunks: 6477
Deduplication factor: 29.90
Average chunk size: 1.426 MiB

The resutlts demonstrate the overhead caused by the additional
ListObjectV2 API calls and their processing, but depending on the
object store backend.

Average garbage collection runtime:
Local datastore:             (2.04 ± 0.01) s
Local RADOS gateway (Squid): (3.05 ± 0.01) s
AWS S3:                      (3.05 ± 0.01) s
Cloudflare R2:               (6.71 ± 0.58) s

After pruning of all datastore contents (therefore including
DeleteObjects requests):
Local datastore:              3.04 s
Local RADOS gateway (Squid): 14.08 s
AWS S3:                      13.06 s
Cloudflare R2:               78.21 s

Signed-off-by: Christian Ebner <c.ebner@proxmox.com>
Reviewed-by: Hannes Laimer <h.laimer@proxmox.com>
Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
2025-07-22 21:43:43 +02:00
..
examples pbs-datastore: clippy fixes 2023-05-17 17:43:17 +02:00
src datastore: implement garbage collection for s3 backend 2025-07-22 21:43:43 +02:00
Cargo.toml datastore: local chunk reader: read chunks based on backend 2025-07-22 21:43:43 +02:00