feat(querybook): use custom iamge to install sqlalchemy-trino

This commit is contained in:
Masaki Yatsu
2025-10-19 20:38:38 +09:00
parent 8d29fe25c0
commit adc009c602
3 changed files with 205 additions and 9 deletions

View File

@@ -54,8 +54,108 @@ QUERYBOOK_NAMESPACE=querybook # Kubernetes namespace
QUERYBOOK_HOST=querybook.example.com # External hostname
KEYCLOAK_HOST=auth.example.com # Keycloak hostname
KEYCLOAK_REALM=buunstack # Keycloak realm name
# Optional: Use custom Docker image (for testing fixes/patches)
QUERYBOOK_CUSTOM_IMAGE=localhost:30500/querybook # Custom image repository
QUERYBOOK_CUSTOM_IMAGE_TAG=trino-metastore # Custom image tag (default: latest)
QUERYBOOK_CUSTOM_IMAGE_PULL_POLICY=Always # Image pull policy (default: Always)
```
### Using Custom Image
To use a custom Querybook image (e.g., with patches or fixes):
```bash
# Set environment variables
export QUERYBOOK_CUSTOM_IMAGE=localhost:30500/querybook
export QUERYBOOK_CUSTOM_IMAGE_TAG=trino-metastore
# Install or upgrade Querybook
just querybook::install
# or
just querybook::upgrade
```
**When to use custom image**:
- Testing bug fixes before they are merged upstream
- Applying patches for specific issues (e.g., datetime JSON serialization)
- Using Trino Metastore integration (requires sqlalchemy-trino)
- Using modified versions with custom features
**Custom image includes** (`trino-metastore` tag):
- Datetime JSON serialization fixes for WebSocket communication
- `sqlalchemy-trino` package for Metastore integration
**Custom image behavior** (when `QUERYBOOK_CUSTOM_IMAGE` is set):
- Pull policy: `Always` (default, override with `QUERYBOOK_CUSTOM_IMAGE_PULL_POLICY`)
- Ensures latest image is always pulled from registry
**Default behavior** (when `QUERYBOOK_CUSTOM_IMAGE` is not set):
- Uses official image: `querybook/querybook:latest`
- Pull policy: `IfNotPresent`
- Note: Official image does not include `sqlalchemy-trino`, so Trino Metastore integration will not work
### Building Custom Image
To build a custom Querybook image with `sqlalchemy-trino` support:
1. **Clone Querybook repository**:
```bash
git clone https://github.com/pinterest/querybook.git
cd querybook
```
2. **Create requirements/local.txt**:
```bash
cat > requirements/local.txt <<EOF
# Local additional requirements for buun-stack
# SQLAlchemy dialect for Trino (required for Metastore)
sqlalchemy-trino
EOF
```
3. **Build the Docker image**:
```bash
# For remote Docker host (e.g., k3s node)
DOCKER_HOST=ssh://yourdomain.com docker build \
--build-arg EXTRA_PIP_INSTALLS=extra.txt \
-t localhost:30500/querybook:trino-metastore .
# For local Docker
docker build \
--build-arg EXTRA_PIP_INSTALLS=extra.txt \
-t localhost:30500/querybook:trino-metastore .
```
4. **Push to registry**:
```bash
DOCKER_HOST=ssh://yourdomain.com docker push localhost:30500/querybook:trino-metastore
# or for local Docker
docker push localhost:30500/querybook:trino-metastore
```
5. **Deploy to Kubernetes**:
```bash
export QUERYBOOK_CUSTOM_IMAGE=localhost:30500/querybook
export QUERYBOOK_CUSTOM_IMAGE_TAG=trino-metastore
just querybook::upgrade
```
**Notes**:
- The Dockerfile automatically includes `requirements/local.txt` if it exists (lines 40-42)
- `EXTRA_PIP_INSTALLS=extra.txt` ensures additional dependencies are installed during build
- The custom image will have both the official Querybook packages and `sqlalchemy-trino`
## Usage
### Access Querybook
@@ -84,27 +184,74 @@ Admin users can:
1. Log in as an admin user
2. Navigate to Admin → Query Engines
3. Click "Add Query Engine"
4. Configure:
4. Configure basic settings:
```plain
Name: Trino
Language: Trino
Executor: Trino (not SqlAlchemy)
Environment: production (or your preferred environment name)
```
5. Navigate to Admin → Environments → [your environment]
6. Add new query engine connection:
5. Configure connection settings:
```plain
Connection String: trino://trino.example.com:443?SSL=true
Connection String: trino://trino.example.com:443/iceberg?SSL=true
Username: admin
Password: [from just trino::admin-password]
Proxy_user_id: (leave empty to use admin username)
```
7. Optional: Configure additional connection parameters:
- **Catalog**: Specify default catalog (e.g., `postgresql` or `iceberg`)
- **Schema**: Specify default schema
- **Proxy_user_id**: Leave empty or set to enable user impersonation
**Important Notes**:
- **Catalog in Connection String**: Include `/iceberg` (or your catalog name) after the port
- With catalog: `trino://host:443/iceberg?SSL=true` → queries work without `iceberg.` prefix
- Without catalog: `trino://host:443?SSL=true` → queries fail with "Catalog 'hive' not found"
- **Proxy_user_id**: Leave empty (defaults to Username field = admin)
- For user impersonation, configure Trino access control separately
6. Optional: Link to Metastore for table autocompletion:
- **Metastore**: Select created Metastore (see Metastore Configuration section below)
- Enables autocomplete for table and column names in query editor
### Configure Metastore (Optional but Recommended)
Metastore enables table/column autocompletion and provides a browsable table catalog.
**Prerequisites**: Custom image with `sqlalchemy-trino` (official image does not include this package)
1. Navigate to Admin → Metastores
2. Click "Create Metastore"
3. Configure:
```plain
Name: Trino Iceberg
Metastore Loader: SqlAlchemyMetastoreLoader
Connection String: trino://admin:[password]@trino.example.com:443/iceberg?SSL=true
Acct Info (Key-Value):
http_scheme = https
Impersonate: OFF (recommended for shared table catalog)
```
**Important Notes**:
- Include authentication in Connection String: `admin:[password]@host`
- Include catalog in Connection String: `/iceberg` after port
- `http_scheme` must be set to `https` in Acct Info
- Keep Impersonate OFF unless you need per-user table filtering
4. Click "Run Task" to sync table metadata
5. Verify in Admin → Metastores that "Last Synced" timestamp is updated
6. Check left sidebar "Tables" for table list
**Scheduled Updates** (recommended):
- Navigate to Admin → Metastores → [your metastore] → Schedule
- Set cron expression: `0 */6 * * *` (sync every 6 hours)
**Usage**:
- **Tables Sidebar**: Browse schemas and tables, view column details
- **Autocomplete**: Type table/column names in query editor, press Tab or Escape
- **Search**: Use search box in Tables sidebar to find tables by name
### User Impersonation
@@ -124,7 +271,7 @@ Querybook connects to Trino as `admin` but executes queries as the logged-in use
## Architecture
```
```plain
External Users
Cloudflare Tunnel (HTTPS)
@@ -244,6 +391,34 @@ kubectl get pods -n querybook
- Check database exists: `just postgres::list-databases | grep querybook`
- Verify secret exists: `kubectl get secret querybook-config-secret -n querybook`
### Metastore Issues
- **Tables sidebar is empty**
- Check Admin → Metastores for "Last Synced" timestamp
- Click "Run Task" to manually sync
- Verify Metastore is linked to Query Engine (Admin → Query Engines → Metastore field)
- Check worker logs: `kubectl logs -n querybook deployment/worker --tail=100 | grep metastore`
- **Error: "Can't load plugin: sqlalchemy.dialects:trino"**
- Official Querybook image does not include `sqlalchemy-trino`
- Use custom image with `QUERYBOOK_CUSTOM_IMAGE_TAG=trino-metastore`
- See "Using Custom Image" section above
- **Error: "Connection.**init**() got an unexpected keyword argument 'password'"**
- Do not use `password` key in Acct Info
- Embed authentication in Connection String: `trino://admin:[password]@host:port/catalog?SSL=true`
- Set `http_scheme = https` in Acct Info
- **Only system.* schemas visible**
- Connection String is missing catalog specification
- Add `/iceberg` (or your catalog) after port: `trino://host:443/iceberg?SSL=true`
- **Autocomplete not working**
- Verify Query Engine has Metastore linked (Admin → Query Engines → Metastore field)
- Refresh DataDoc page (F5) after linking Metastore
- Check Environment matches between DataDoc and Query Engine
- Try Tab or Escape key instead of Ctrl+Space (macOS shortcut conflict)
## References
- [Querybook Documentation](https://www.querybook.org/)