wiki: add Tailscale SSH reauth article; update Netdata Docker alarm tuning (50 articles)

- New: Tailscale SSH unexpected re-authentication prompt — diagnosis and fix
- Updated: netdata-docker-health-alarm-tuning — add delay: up 3m to suppress
  Nextcloud AIO PHP-FPM ~90s startup false alerts; update settings table and notes
- Updated: 05-troubleshooting/index.md and SUMMARY.md

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
This commit is contained in:
2026-03-21 00:12:52 -04:00
parent 4d59856c1e
commit c4d3f8e974
4 changed files with 74 additions and 5 deletions

View File

@@ -0,0 +1,66 @@
# Tailscale SSH: Unexpected Re-Authentication Prompt
If a Tailscale SSH connection unexpectedly presents a browser authentication URL mid-session, the first instinct is to check the ACL policy. However, this is often a one-off Tailscale hiccup rather than a misconfiguration.
## Symptoms
- SSH connection to a fleet node displays a Tailscale auth URL:
```
To authenticate, visit: https://login.tailscale.com/a/xxxxxxxx
```
- The prompt appears even though the node worked fine previously
- Other nodes in the fleet connect without prompting
## What Causes It
Tailscale SSH supports two ACL `action` values:
| Action | Behavior |
|---|---|
| `accept` | Trusts Tailscale identity — no additional auth required |
| `check` | Requires periodic browser-based re-authentication |
If `action: "check"` is set, every session (or after token expiry) will prompt for browser auth. However, even with `action: "accept"`, a one-off prompt can appear due to a Tailscale daemon glitch or key refresh event.
## How to Diagnose
### 1. Verify the ACL policy
In the Tailscale admin console (or via `tailscale debug acl`), inspect the SSH rules. For a trusted homelab fleet, the rule should use `accept`:
```json
{
"src": ["autogroup:member"],
"dst": ["autogroup:self"],
"users": ["autogroup:nonroot", "root"],
"action": "accept",
}
```
If `action` is `check`, that is the root cause — change it to `accept` for trusted source/destination pairs.
### 2. Confirm it was a one-off
If the ACL already shows `accept`, the prompt was transient. Test with:
```bash
ssh <hostname> "echo ok"
```
No auth prompt + `ok` output = resolved. Note that this test is only meaningful if the previous session's auth token has expired, or you test from a different device that hasn't recently authenticated.
## Fix
**If ACL shows `check`:** Change to `accept` in the Tailscale admin console under Access Controls. Takes effect immediately — no server changes needed.
**If ACL already shows `accept`:** No action required. The prompt was a one-off Tailscale event (daemon restart, key refresh, etc.). Monitor for recurrence.
## Notes
- Port 2222 on **MajorRig** exists as a hard bypass for Tailscale SSH browser auth — regular SSH over Tailscale network, bypassing Tailscale SSH entirely. This is an alternative approach if `check` mode is required for compliance but browser auth is too disruptive.
- The `autogroup:self` destination means the rule applies when connecting from your own devices to your own devices — appropriate for a personal homelab fleet.
## Related
- [[Network Overview]] — Tailscale fleet inventory and SSH access model
- [[SSH-Aliases]] — Fleet SSH access shortcuts