philipp/incus-backup-ui
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

security hardening

Browse Source
This commit is contained in:
Philipp
2026-05-21 14:09:08 +02:00
parent 3103aba972
commit 0046156e58
16 changed files with 205 additions and 43 deletions
Download Patch File Download Diff File Expand all files Collapse all files
README.md
+5 -3
View File
@@ -22,7 +22,7 @@ npm run dev
The node agent must run on every Incus host with permission to access Incus, ZFS, `/dev/zvol`, Restic, and S3 credentials. In production this usually means running it as root or through a tightly scoped service account with the needed privileges.
Set `API_TOKEN` in `backend/.env`; the management server uses that token when calling the agent.
Set `API_TOKEN` in `backend/.env`; the management server uses that token when calling the agent. The token is required and must be at least 32 characters long.
Required commands:
@@ -42,7 +42,9 @@ npm install
npm run dev
```
The management API stores nodes, users, sessions, and central schedules in SQLite. Configure the first admin user through `AUTH_USERNAME` and `AUTH_PASSWORD` before the first start. If no password is configured, the development fallback is `admin`.
The management API stores nodes, users, sessions, and central schedules in SQLite. Configure the first admin user through `AUTH_USERNAME` and `AUTH_PASSWORD` before the first start. Startup fails if the initial password is missing.
Agent URLs must use `https://` by default. For local development only, set `ALLOW_INSECURE_AGENT_HTTP=true` in `management/.env` to permit `http://` node URLs.
The management API uses Node's built-in SQLite module and requires Node.js 22.5 or newer.
@@ -71,7 +73,7 @@ Changing most node-agent values applies to new API calls and jobs immediately. C
## Safety Notes
Restore is intentionally guarded twice: the backend validates the snapshot against the VM, and the UI requires typing the VM name before sending the restore request. Restore jobs are never retried automatically.
Restore is intentionally guarded twice: the backend validates the snapshot against the VM, and the UI requires typing the VM name before sending the restore request. VM restores create a pre-restore ZFS snapshot and roll back to it if writing the disk fails. Restore jobs are never retried automatically.
## Deployment