How to Self-Host Plausible Analytics With Docker Compose
Plausible Community Edition (CE) is the self-hosted version of Plausible, a web analytics service. Docker Compose uses a configuration file to run…
Categories
Is S3 Glacier Deep Archive a Good Off-Site Backup for a Home Server?
If you back up a home server to Amazon S3, Glacier Deep Archive can keep an off-site copy at a very low storage price. An off-site backup is a copy stored away from your home, so a fire, theft, or other local disaster is less likely to destroy both the server and its backup.
Deep Archive is an archival storage tier: you must request a restore before you can read an archived file. That delay makes it a good fit for data you rarely need and a poor choice as your only backup when you need to get a server running quickly. For many home setups, keep a faster local backup and use Deep Archive for an additional copy that you rarely need to access.
What recovery from Deep Archive involves
Amazon S3 Glacier Deep Archive is a storage class, a pricing and access tier that sets how S3 charges for and lets you access items in a bucket. S3 stores each item as an object, which includes the data and its metadata. A separate backup tool or workflow must create the objects and manage retention (how long to keep them). It must also restore them. Deep Archive is also separate from the older Amazon Glacier vault service, which is a different service.
To recover an archived object, you first submit a restore request. S3 then makes a temporary copy available for the period you specify. After the restore finishes, you download the temporary copy. S3 charges for storing it separately from the archived copy. The S3 archival-storage documentation describes this two-stage path.
AWS describes Deep Archive Standard retrieval as typically taking up to 12 hours, with 9 to 12 hours cited for S3 Batch Operations. Bulk retrieval typically takes up to 48 hours. AWS also describes Deep Archive as suitable when a user can wait up to 72 hours. AWS's times are typical service estimates, so they do not set a guaranteed deadline for complete home-server recovery. After S3 restores the data, your home internet connection must download it. A large backup or a backup split into many objects can take longer to recover end to end.
What a restore costs
Deep Archive’s advertised storage price is $0.00099 per GB-month, or about $1 per TB-month. That makes it inexpensive to keep data for a long time. A recovery also incurs charges for data retrieval and S3 requests. AWS charges separately to store the restored copy temporarily, and internet data transfer can add to the bill.
For a regional example, AWS’s published rates for US East (N. Virginia) list retrieval at $0.02 per GB for Standard and $0.0025 per GB for Bulk. Retrieving 1 TiB, or 1,024 GB for this calculation, would cost about $20.48 with Standard or $2.56 with Bulk. Neither estimate includes request charges or temporary storage. Internet transfer is also excluded. AWS’s pricing example is specific to that region, so use the current pricing for your bucket’s region to estimate your own bill.
Object count matters as well as backup size. AWS bills extra metadata for each Deep Archive object, and request charges rise with the number of objects. A backup made of many small files can therefore cost more to transition and manage than a backup packaged into fewer, larger objects. Deep Archive also has a 180-day minimum storage duration. If you delete, overwrite, or transition an object sooner, AWS can charge for the remaining portion of that minimum period. Check the S3 storage-class details and current S3 pricing before setting retention rules.
When Deep Archive fits a home server
Deep Archive fits a backup copy intended for rare emergencies, such as recovering after a fire, theft, or major hardware failure. It makes sense when you can wait a day or more before retrieving the data, expect to retain each archived copy for at least 180 days, and can manage the bucket, credentials, encryption keys, and restore steps.
Deep Archive is a poor fit as the only backup for a server that needs quick recovery. Frequent file restores also make it a poor fit, as does backup software that expects every stored object to be immediately readable. Compatibility depends on the backup tool and how it stores data. For example, the FAQ for restic 0.18.0 describes its Glacier and Deep Archive restore support as experimental. That statement applies to that specific version, so check the documentation for the exact tool and version you use.
For a faster cloud recovery path, consider S3 Glacier Flexible Retrieval. AWS describes retrieval options that range from minutes to hours and recommends the class for some backup and disaster-recovery uses. It costs more to store than Deep Archive, but it gives you more flexibility when a restore cannot wait a day or longer.
Hot object storage offers another trade-off: it costs more each month but keeps objects readily accessible. Backblaze B2 lists a starting storage price of $6.95 per TB per 30 days. It offers free egress up to three times average monthly stored data and charges for additional egress. Wasabi lists $7.99 per TB-month for the US and Europe. It also has a 1 TB monthly minimum and a 90-day minimum storage duration. Review the providers’ Backblaze pricing and Wasabi product terms, because rates and conditions can change.
| Option | Best fit | Trade-off |
|---|---|---|
| S3 Glacier Deep Archive | A rarely accessed, long-term copy | Low storage cost, but restore requests and downloads take time and add charges |
| S3 Glacier Flexible Retrieval | A cloud backup that needs more retrieval flexibility | Higher storage cost than Deep Archive |
| Backblaze B2 or Wasabi | Readily accessible cloud copies | Higher monthly storage cost and provider-specific egress or minimum-use terms |
| A second off-site drive or server | A familiar recovery path without archive-tier retrieval | Requires a secure location and a way to keep the copy current |
Test the recovery path before relying on it
A successful backup upload does not prove that you can recover the server. Run a test restore that follows the same steps you would use after a failure: request the archive restore, then retrieve the objects. Use your backup tool to restore files or data. Record how long each step takes over your actual home connection, and confirm that you still have the credentials and encryption keys needed to read the backup.
To estimate the cost of a full recovery, use your S3 region, the total amount of data, the number of objects, the retrieval tier, how long S3 will store the temporary copy, and the expected internet data transfer. If ransomware or accidental deletion is a concern, S3 Object Lock can add retention controls that prevent protected object versions from being deleted during the retention period. Object Lock requires versioning, and its settings need to match your retention and recovery plan.
What I'm buildingDelegate tasks. Get software.
Give Vroni a GitHub issue, bug report, spec, or rough idea. It reads the repo, plans the change, writes code, runs checks, and works toward a review-ready pull request.
Take a look at vroni.com