Host Troubleshooting Playbook
When your host has issues, the last thing you want is to dig through docs. This page lists rescue steps by symptom — find the right action in 30 seconds max.
Don’t immediately “Restart Server”. Problems that restart can’t fix will still be there after restart. Diagnose by symptom first, then act.
Symptom 1: admin-panel won’t open at all
What you see
- Browser shows “Cannot connect” / “Connection timeout” / “DNS not found”
- Spinning for several minutes
- Even Magic Link doesn’t help
1-min checks
| Check | How |
|---|---|
| 1. Your network | Open another site (google.com) — works? Your network is OK |
| 2. realvco service status | Open https://realvco.com — works? realvco is OK overall |
| 3. Host URL correct | Verify the URL in your setup email is https://your-host-00.realvco.com/ |
Action
- 2 and 3 OK but still down → Host itself may be down. Submit urgent ticket, title with “Host completely unresponsive”, realvco support will check host health
- 2 down but google works → realvco service-wide outage, wait or check realvco status
- Network completely broken → Fix your network
Symptom 2: admin-panel loads but AI agents show red lights
What you see
- admin-panel HOST tab loads
- But all 3 agents in AI Agent List show red / gray
- Click into OC-1 / OC-2 / HM-3 shows “container not running”
1-min checks
Go to Host Operations and look at:
- Is SYSTEM UPTIME very short (< 5 min)? → Host just restarted, wait for containers to auto-start (5-10 min)
- Is it very long (> 30 days)? → Possibly progressive degradation
Action
| Situation | Action |
|---|---|
| Host just restarted | Wait 10 min for containers to auto-start |
| Still red after 10 min | Restart Server (red button) |
| Still red after restart | Submit ticket with hostname + last working time |
⚠️ Don’t restart 3+ times — if 3 restarts fail it’s a real issue, ticket is faster.
Symptom 3: One AI agent unresponsive (others fine)
What you see
- OC-1 (Rose) replies on Telegram ✅
- OC-2 (Ada) doesn’t reply ❌
- HM-3 (Vi) also fine
Only 1 agent broken — no need to touch host, just restart that single AI.
Action
- Dashboard → that AI agent tab (e.g. OC-2) → Home
- Click Pause button → wait 5 sec
- Click Resume button
- Wait 30 sec for container to start
- Re-test by sending message on Telegram
90% of single-AI hangs solve via “Pause → Resume”.
If “Pause → Resume” doesn’t work:
- Check Activity sub-tab for error log
- Check Usage sub-tab for sudden token spike (might have run out)
Symptom 4: AI behaves weirdly after upgrade / amnesia
What you see
- Just upgraded OC-1 version
- AI doesn’t recognize itself (persona seems lost)
- Or AI memory broken (forgets previous conversations)
Action — One-click rollback within 5 days
- AI agent → Settings → Local Backup Versions
- Click ”↶ Restore from Local Backup [old version] (X days left)”
- Wait 5-10 min, AI back to old version
⚠️ Local backup auto-clears after 5 days, don’t drag failed upgrades.
Symptom 5: Storage / CPU / RAM 100%
What you see
- admin-panel loads slowly
- AI replies slowly
- Host Info shows disk > 95%
Action
By risk, low to high:
| # | Action | Risk |
|---|---|---|
| 1 | Free Up Space (clean Docker cache) | 0 — clears unused images |
| 2 | Export conversation history then delete old | Low — exported data preserved |
| 3 | Restart host | Medium — short downtime |
| 4 | Upgrade host plan (Contabo add disk) | High — paid + maintenance window |
90% of “disk full” is solved by Free Up Space — Docker cumulative images are biggest culprit.
Symptom 6: Everything fine but token usage spiked
What you see
- Normally X token / day, today used 5X
- But you didn’t do anything special
- AI seems to be replying normally
1-min checks
Go to that AI agent → Activity sub-tab filter “Running” “Failed”:
- Are there repeating same tasks? → Possibly cron in retry loop
- Are there very long conversations? → A customer might be exhausting context
- Are there strange conversation sources? → Possibly token leaked, someone misusing AI
Action
| Situation | Action |
|---|---|
| Cron retry loop | In Activity sub-tab find that schedule, pause it |
| Long customer conversation | Check if reasonable; if not, add to personality “after N lines escalate” |
| Suspected token leak | Immediately pause that AI + rotate all platform tokens + submit ticket |
Symptom 7: Completely lost
| Stuck type | Direct action |
|---|---|
| Haven’t logged in, don’t know where to start | Steps to master AI agent |
| Failed upgrade can’t find rollback button | Local Backup Versions |
| Want to stop all AI but afraid to delete | All 3 agents click Pause (not delete) |
| All methods above tried, still broken | Submit ticket, with hostname + what you tried |
Prevention beats cure
| Action | Frequency | Why |
|---|---|---|
| Create Restore Point | Before upgrade / major changes | One-click rollback |
| Backup Now + Download | Monthly | Conversations + settings offline backup |
| Check Usage sub-tab | Weekly | Catch token anomalies early |
| Check Activity sub-tab | Weekly | Catch AI error patterns early |
Still not working?
Last resort: Submit ticket.
Ticket should include:
- ☐ Hostname (e.g.
cattle42) - ☐ Symptom (be specific: “OC-1 OpenClaw UI returns 404” beats “Rose broken”)
- ☐ When it started
- ☐ What you’ve tried (use this page’s numbering, e.g. “Symptom 2 + tried Pause/Resume 3 times”)
- ☐ Urgency (business waiting / can wait until tomorrow)
realvco support typically responds within 1-2 hours.