Part of #2. Gitea is about to carry builds and deployments, and nothing automated backs up its repositories, database, or packages.
Decision (Eric): the backup destination and retention. An off-site location is preferred, per #2. Work (Claude): add a scheduled, encrypted backup to the role, with an alert on failure or missed runs.
Acceptance: backups run on schedule and land at the destination. A failure raises an alert.
Label ready-to-deploy when committed.
Part of #2. Gitea is about to carry builds and deployments, and nothing automated backs up its repositories, database, or packages.
**Decision (Eric)**: the backup destination and retention. An off-site location is preferred, per #2.
**Work (Claude)**: add a scheduled, encrypted backup to the role, with an alert on failure or missed runs.
**Acceptance**: backups run on schedule and land at the destination. A failure raises an alert.
Label `ready-to-deploy` when committed.
claude
added this to the CI/CD 4: Continuous delivery milestone 2026-11-09 17:07:09 +00:00
Estimate: about $10–25 a month to back up the whole platform to AWS, plus roughly $100–500 if you ever do a full restore. Workings below.
What there is to back up (measured read-only today from disk usage on the two NFS servers, not from PVC sizes, which are much larger than what's used)
Set
Size
Back up?
Platform data: all cluster volumes (object storage about 0.94 TB, which is most of it; PeerTube about 80 GB; Gitea, databases, www, Nextcloud app data and the rest are each a few GB)
about 1.1 TB
Yes, the core of the plan
Your personal files on the main NFS server
about 4.3 TiB (4.7 TB)
Your call. Mostly large files that rarely change
AI model files, blockchain node data, archived-* volumes
about 0.45 TB
No. They can be downloaded or rebuilt again
VM and node root disks
about 0.85 TB
No. They're rebuilt from this repo with Ansible
Monthly storage cost (AWS us-east-1 list prices as I know them from mid-2026. I didn't check them live, so confirm in the AWS Pricing Calculator)
S3 class
per TB-month
Platform, 1.1 TB
Personal, 4.7 TB
Both
Standard
about $23
$25
$108
$133
Standard-IA
about $12.50
$14
$59
$73
Glacier Instant Retrieval
about $4
$4.40
$19
$23
Glacier Deep Archive
about $1
$1.10
$4.70
$6
Add 10–30% for retention history (daily snapshots kept 30 days, then monthly), depending on how fast data changes.
One-off and occasional costs
Upload: transfer into AWS is free. Request charges are small if the backup tool packs files into large objects: roughly $2–20 for the first full upload, depending on class. The time depends on your upload speed. 5.8 TB takes about 13 hours at 1 Gbps or about 11 days at 50 Mbps.
Restore: the main cost. Transfer out is about $0.09/GB, so a full platform restore (1.1 TB) is about $90–100, and everything (5.8 TB) is about $500, plus a retrieval fee: about $0.03/GB for Glacier Instant Retrieval, about $0.0025/GB for Deep Archive bulk (which takes up to 48 hours). Restoring a single file costs cents.
Minimum storage periods: 30 days for Standard-IA, 90 for Glacier Instant Retrieval, 180 for Deep Archive. Deleting earlier is billed as if kept.
Recommendation: about $8–12 a month
Platform data → restic → S3 Glacier Instant Retrieval, nightly, encrypted on our side before upload. Restores are immediate. Databases are dumped with pg_dump first rather than copying live database files, so the backup is consistent. About $5 a month.
Personal files → rclone → S3 Glacier Deep Archive, weekly, with client-side encryption. Cheapest for large files that rarely change. A restore takes up to two days, which suits a disaster copy. About $5 a month.
A dedicated bucket with versioning and Object Lock, so a compromised host can't delete backups, and an IAM user that can only write to that bucket. Restores use separate credentials that you keep. Retention is set with a lifecycle rule.
A failure or missed-run alert in Prometheus, as this issue already plans.
What it needs from you
Go/no-go on AWS as the off-site destination, and whether personal files are included.
Creating the bucket and the backup-only IAM user (accounts and credentials are yours to create). I won't use my own AWS access for this. It's limited to email.
Roughly your upload speed, so I can set the schedule and bandwidth cap.
Once you decide, I'll write the Gitea part of D1 as the first piece: the role change, the encrypted schedule, the alert, and ready-to-deploy. I'll structure it so the other services can reuse the same pattern under #2. needs-eric stays for the destination decision.
— eom-cicd-roadmap
**Estimate: about $10–25 a month to back up the whole platform to AWS, plus roughly $100–500 if you ever do a full restore.** Workings below.
**What there is to back up** (measured read-only today from disk usage on the two NFS servers, not from PVC sizes, which are much larger than what's used)
| Set | Size | Back up? |
|---|---|---|
| Platform data: all cluster volumes (object storage about 0.94 TB, which is most of it; PeerTube about 80 GB; Gitea, databases, www, Nextcloud app data and the rest are each a few GB) | **about 1.1 TB** | Yes, the core of the plan |
| Your personal files on the main NFS server | **about 4.3 TiB** (4.7 TB) | Your call. Mostly large files that rarely change |
| AI model files, blockchain node data, `archived-*` volumes | about 0.45 TB | No. They can be downloaded or rebuilt again |
| VM and node root disks | about 0.85 TB | No. They're rebuilt from this repo with Ansible |
**Monthly storage cost** (AWS us-east-1 list prices as I know them from mid-2026. I didn't check them live, so confirm in the AWS Pricing Calculator)
| S3 class | per TB-month | Platform, 1.1 TB | Personal, 4.7 TB | Both |
|---|---|---|---|---|
| Standard | about $23 | $25 | $108 | $133 |
| Standard-IA | about $12.50 | $14 | $59 | $73 |
| Glacier Instant Retrieval | about $4 | $4.40 | $19 | $23 |
| Glacier Deep Archive | about $1 | $1.10 | $4.70 | $6 |
Add 10–30% for retention history (daily snapshots kept 30 days, then monthly), depending on how fast data changes.
**One-off and occasional costs**
- **Upload:** transfer into AWS is free. Request charges are small if the backup tool packs files into large objects: roughly $2–20 for the first full upload, depending on class. The time depends on your upload speed. 5.8 TB takes about 13 hours at 1 Gbps or about 11 days at 50 Mbps.
- **Restore:** the main cost. Transfer out is about $0.09/GB, so a full platform restore (1.1 TB) is about $90–100, and everything (5.8 TB) is about $500, plus a retrieval fee: about $0.03/GB for Glacier Instant Retrieval, about $0.0025/GB for Deep Archive bulk (which takes up to 48 hours). Restoring a single file costs cents.
- **Minimum storage periods:** 30 days for Standard-IA, 90 for Glacier Instant Retrieval, 180 for Deep Archive. Deleting earlier is billed as if kept.
**Recommendation: about $8–12 a month**
1. **Platform data → `restic` → S3 Glacier Instant Retrieval**, nightly, encrypted on our side before upload. Restores are immediate. Databases are dumped with `pg_dump` first rather than copying live database files, so the backup is consistent. About $5 a month.
2. **Personal files → `rclone` → S3 Glacier Deep Archive**, weekly, with client-side encryption. Cheapest for large files that rarely change. A restore takes up to two days, which suits a disaster copy. About $5 a month.
3. A dedicated bucket with versioning and Object Lock, so a compromised host can't delete backups, and an IAM user that can only write to that bucket. Restores use separate credentials that you keep. Retention is set with a lifecycle rule.
4. A failure or missed-run alert in Prometheus, as this issue already plans.
**What it needs from you**
- **Go/no-go on AWS** as the off-site destination, and whether personal files are included.
- Creating the bucket and the backup-only IAM user (accounts and credentials are yours to create). I won't use my own AWS access for this. It's limited to email.
- Roughly your upload speed, so I can set the schedule and bandwidth cap.
Once you decide, I'll write the Gitea part of D1 as the first piece: the role change, the encrypted schedule, the alert, and `ready-to-deploy`. I'll structure it so the other services can reuse the same pattern under #2. `needs-eric` stays for the destination decision.
— eom-cicd-roadmap
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
Part of #2. Gitea is about to carry builds and deployments, and nothing automated backs up its repositories, database, or packages.
Decision (Eric): the backup destination and retention. An off-site location is preferred, per #2.
Work (Claude): add a scheduled, encrypted backup to the role, with an alert on failure or missed runs.
Acceptance: backups run on schedule and land at the destination. A failure raises an alert.
Label
ready-to-deploywhen committed.I actually need to establish a backup plan for my entire platform. Can you give me an estimate of how much it would cost to do this with AWS?
Estimate: about $10–25 a month to back up the whole platform to AWS, plus roughly $100–500 if you ever do a full restore. Workings below.
What there is to back up (measured read-only today from disk usage on the two NFS servers, not from PVC sizes, which are much larger than what's used)
archived-*volumesMonthly storage cost (AWS us-east-1 list prices as I know them from mid-2026. I didn't check them live, so confirm in the AWS Pricing Calculator)
Add 10–30% for retention history (daily snapshots kept 30 days, then monthly), depending on how fast data changes.
One-off and occasional costs
Recommendation: about $8–12 a month
restic→ S3 Glacier Instant Retrieval, nightly, encrypted on our side before upload. Restores are immediate. Databases are dumped withpg_dumpfirst rather than copying live database files, so the backup is consistent. About $5 a month.rclone→ S3 Glacier Deep Archive, weekly, with client-side encryption. Cheapest for large files that rarely change. A restore takes up to two days, which suits a disaster copy. About $5 a month.What it needs from you
Once you decide, I'll write the Gitea part of D1 as the first piece: the role change, the encrypted schedule, the alert, and
ready-to-deploy. I'll structure it so the other services can reuse the same pattern under #2.needs-ericstays for the destination decision.— eom-cicd-roadmap