Document current architecture, manual fixes, and QA checklist

Added comprehensive documentation:
- Manual workaround for sender_localpart registration bug
- QA testing checklist for untested features
- Future monitoring/alerting requirements
- Current architecture diagram and data flow
- Security model and operational notes
This commit is contained in:
Dan 2025-10-26 14:52:31 -07:00
parent 0b1751766b
commit 2dfe4ea829

266
CLAUDE.md
View file

@ -131,14 +131,16 @@ ssh root@45.77.205.49 'sudo -u postgres pg_dump mautrix_slack' > backup.sql
## Recent Changes ## Recent Changes
- 001-extract-matrix-platform: Added Nix 2.x, NixOS 24.05+, Bash 5.x (for scripts) - 001-extract-matrix-platform: Added Nix 2.x, NixOS 24.05+, Bash 5.x (for scripts)
- 002-slack-bridge-integration: Added mautrix-slack bridge with Socket Mode (2025-10-22) - 002-slack-bridge-integration: Deployed mautrix-slack bridge with Socket Mode (2025-10-26)
- Phase 0: Research complete (Socket Mode, API scopes, sops-nix, channel patterns) - Phase 0-1: Research and design complete
- Phase 1: Design complete (data-model.md, contracts/, quickstart.md) - Phase 2: Infrastructure deployed and operational
- Next: Phase 2 (/speckit.tasks for implementation breakdown) - Status: Bidirectional message flow working (Slack ↔ Matrix)
- ~50 Slack channels synced to Matrix rooms
## Known Issues ## Known Issues
- mautrix-slack exits with code 11 without credentials (expected, requires `login app`)
- olm-3.2.16 marked insecure (permitted via nixpkgs.config.permittedInsecurePackages) - olm-3.2.16 marked insecure (permitted via nixpkgs.config.permittedInsecurePackages)
- conduwuit log level set to "debug" (intended for troubleshooting, consider reverting to "info")
- Fresh database required after conduwuit version upgrades (wipe /var/lib/matrix-continuwuity/db/)
## Testing Guidelines ## Testing Guidelines
- Test message latency: Should be <5 seconds (FR-001, FR-002) - Test message latency: Should be <5 seconds (FR-001, FR-002)
@ -147,4 +149,258 @@ ssh root@45.77.205.49 'sudo -u postgres pg_dump mautrix_slack' > backup.sql
- Stability target: 99% uptime over 7-day period - Stability target: 99% uptime over 7-day period
<!-- MANUAL ADDITIONS START --> <!-- MANUAL ADDITIONS START -->
## Manual Configuration Workarounds
### mautrix-slack Registration File Fix (KNOWN ISSUE)
**Problem:** The bridge's registration generator creates a random `sender_localpart` instead of using the configured `bot.username` value.
**Current Manual Fix (Required on Fresh Deploy):**
```bash
# After bridge service starts and generates registration
ssh root@45.77.205.49 'systemctl stop mautrix-slack'
# Edit registration file to fix sender_localpart
ssh root@45.77.205.49 "sed -i 's/^sender_localpart: .*/sender_localpart: slackbot/' /var/lib/matrix-appservices/mautrix_slack_registration.yaml"
# Re-register appservice in Matrix admin room
# In Element, send to admin room:
# !admin appservices unregister slack
# !admin appservices register
# <paste corrected YAML>
# Restart homeserver to load new registration
ssh root@45.77.205.49 'systemctl restart matrix-continuwuity'
# Start bridge
ssh root@45.77.205.49 'systemctl start mautrix-slack'
```
**Root Cause:** mautrix-slack's `-g` flag generates registration independently of `config.yaml` settings.
**Potential Permanent Fix:** Patch `modules/mautrix-slack.nix` to post-process registration file after generation:
```nix
# In ExecStartPre, after registration generation:
${pkgs.gnused}/bin/sed -i 's/^sender_localpart: .*/sender_localpart: ${cfg.appservice.senderLocalpart}/' "$REG_PATH"
```
**Impact:** Without this fix, registration sender_localpart won't match bridge config, causing authentication failures.
---
## QA Testing Checklist
### Core Features (✅ Tested & Working)
- [x] Bidirectional text messaging (Slack ↔ Matrix)
- [x] Channel discovery and room creation (~50 channels synced)
- [x] Socket Mode WebSocket connection
- [x] Bot authentication with Matrix homeserver
- [x] Bridge startup and recovery after restart
### Features Requiring QA Testing (⚠️ Untested)
- [ ] **File Attachments**
- Upload file in Slack → verify appears in Matrix
- Upload file in Matrix → verify appears in Slack
- Test various file types (images, PDFs, archives)
- Test large files (>10MB)
- [ ] **Emoji Reactions**
- Add reaction in Slack → verify appears in Matrix
- Add reaction in Matrix → verify appears in Slack
- Remove reaction → verify syncs
- [ ] **Message Edits**
- Edit message in Slack → verify updates in Matrix
- Edit message in Matrix → verify updates in Slack
- [ ] **Message Deletion**
- Delete message in Slack → verify removes from Matrix
- Delete message in Matrix → verify removes from Slack
- [ ] **Thread Replies**
- Reply in Slack thread → verify threading in Matrix
- Reply in Matrix thread → verify threading in Slack
- [ ] **User Profile Sync**
- Change Slack display name → verify updates Matrix puppet
- Change Slack avatar → verify updates Matrix puppet
- [ ] **Error Handling**
- Network interruption recovery
- Matrix homeserver restart handling
- Slack WebSocket reconnection
- Invalid token handling
- [ ] **Performance**
- High-volume channel (>100 messages/hour)
- Large file transfer times
- Message latency under load
### Test Commands
```bash
# Monitor bridge during testing
ssh root@45.77.205.49 'journalctl -u mautrix-slack -f'
# Check for errors
ssh root@45.77.205.49 'journalctl -u mautrix-slack --since "1 hour ago" | grep -E "ERR|WRN|FTL"'
# Verify message flow
# Test in #vlads-pad or similar channel
# Send from Slack, verify in Matrix room
# Send from Matrix room, verify in Slack
```
---
## Future Infrastructure Needs
### Monitoring & Alerting (Not Implemented)
**Health Checks Needed:**
- Bridge WebSocket connection status
- Matrix homeserver availability
- Message processing latency
- Database connection health
- Error rate thresholds
**Potential Solutions:**
```bash
# Option 1: Simple systemd monitoring
systemctl status mautrix-slack | grep -q "active (running)" || alert
# Option 2: Prometheus + Alertmanager
# - Export bridge metrics (if available)
# - Alert on service down, high error rate, message lag
# Option 3: Uptime monitoring
# - External ping to Matrix homeserver
# - Check /_matrix/client/versions endpoint
# - Alert on HTTP errors or timeout
```
**Metrics to Track:**
- Bridge uptime percentage
- Messages processed (Slack → Matrix, Matrix → Slack)
- WebSocket reconnection events
- Database query performance
- Error counts by type
**Alert Conditions:**
- Bridge down for >5 minutes
- No messages processed in >15 minutes (if active channels exist)
- Error rate >5% of total messages
- Database connection failures
- Disk space <10% free
### Backup Strategy (Not Implemented)
**Critical Data:**
- Matrix RocksDB: `/var/lib/matrix-continuwuity/db/` (66M)
- Bridge PostgreSQL: `mautrix_slack` database (172K)
- Registration files: `/var/lib/matrix-appservices/*.yaml`
- Secrets: sops-encrypted `secrets/secrets.yaml` (in git)
**Backup Approach:**
```bash
# Daily database backups
ssh root@45.77.205.49 'tar czf /root/backups/matrix-$(date +%Y%m%d).tar.gz /var/lib/matrix-continuwuity/db/'
ssh root@45.77.205.49 'sudo -u postgres pg_dump mautrix_slack > /root/backups/bridge-$(date +%Y%m%d).sql'
# Retention: 7 daily, 4 weekly, 12 monthly
# Store off-VPS (rsync to backup server or cloud storage)
```
**Recovery Procedure:**
1. Deploy NixOS configuration
2. Restore database backups
3. Restore registration files
4. Re-authenticate with Slack (new tokens via `login app`)
5. Verify message flow
**Note:** Matrix database can be wiped and rebuilt from Slack if needed (current architecture treats Matrix as ephemeral view layer).
---
## Current Architecture State (2025-10-26)
### Deployed Services
```
┌─────────────────────────────────────────────────────┐
│ clarun.xyz (45.77.205.49) │
│ │
│ ┌─────────────────────────────────────────────┐ │
│ │ nginx :443 (HTTPS) │ │
│ │ - Matrix Client-Server API │ │
│ │ - Forgejo (git.clarun.xyz) │ │
│ └────────────┬────────────────────────────────┘ │
│ │ │
│ ├─→ conduwuit :8008 (127.0.0.1) │
│ │ - Matrix homeserver │
│ │ - RocksDB schema v18 │
│ │ - 66M database │
│ │ │
│ └─→ Forgejo :3000 (127.0.0.1) │
│ │
│ ┌─────────────────────────────────────────────┐ │
│ │ mautrix-slack :29319 (127.0.0.1) │ │
│ │ - Socket Mode WebSocket to Slack │ │
│ │ - PostgreSQL backend (172K) │ │
│ │ - ~50 portal rooms │ │
│ └────────────┬────────────────────────────────┘ │
│ │ │
│ └─→ PostgreSQL :5432 (unix socket) │
│ │
└─────────────────────────────────────────────────────┘
└─→ Slack API (Socket Mode WebSocket)
- Workspace: chochacho
- Bot token: xoxb-...
- App token: xapp-...
```
### Critical Networking Details
- **All internal services use IPv4 (127.0.0.1)** - NOT "localhost"
- Reason: `localhost` resolves to IPv6 `[::1]` but services bind IPv4-only
- Fixed in: nginx proxy_pass, bridge homeserverUrl configuration
### Service Dependencies
```
postgresql.service
└─→ mautrix-slack.service
└─→ matrix-continuwuity.service
└─→ nginx.service
```
### Data Flow
1. **Slack → Matrix:**
- Slack pushes event via Socket Mode WebSocket
- Bridge receives, transforms to Matrix event
- Bridge POSTs to conduwuit appservice endpoint
- conduwuit distributes to Matrix rooms
- Element clients receive via /sync
2. **Matrix → Slack:**
- Element client sends message via conduwuit
- conduwuit forwards to bridge appservice endpoint
- Bridge transforms to Slack API call
- Bridge POSTs to Slack API (bot token)
- Appears in Slack channel
### Security Model
- **Secrets:** Managed via sops-nix, deployed to `/run/secrets/`
- **Bridge tokens:**
- `as_token`: Bridge authenticates to Matrix
- `hs_token`: Matrix authenticates to bridge
- **Slack tokens:**
- `xoxb-`: Bot API calls
- `xapp-`: Socket Mode connection
- **No public bridge endpoint:** Socket Mode eliminates webhook requirement
### Operational Notes
- Matrix database disposable (can rebuild from Slack)
- Bridge config fully declarative except sender_localpart fix
- Fresh database recommended after conduwuit version upgrades
- Debug logging currently enabled on conduwuit
<!-- MANUAL ADDITIONS END --> <!-- MANUAL ADDITIONS END -->