Migrate Mquery to typed config library (#324)

This changes the current mquery configuration mechanism from editing config.py file, to a "real" config file like this (mquery.ini):

```ini
[redis]
host=redis-server.example.com

[mquery]
backend=tcp://ursadb-server.example.com:9281
plugins=plugins.archive:GzipPlugin
```

.github/workflows/test_code.yml

@@ -28,8 +28,6 @@ jobs:
         run: pip3 install mypy==0.790
       - name: install requirements
         run: pip3 install -r requirements.txt
-      - name: copy config
-        run: cp src/config.example.py src/config.py
       - name: run mypy
         run: mypy src
   test_python_style:
@@ -41,8 +39,6 @@ jobs:
         uses: actions/setup-python@v1
         with:
           python-version: '3.10'
-      - name: copy config
-        run: cp src/config.example.py src/config.py
       - name: install flake8==3.7.9
         run: pip3 install flake8==3.7.9
       - name: run flake8
@@ -56,8 +52,6 @@ jobs:
         uses: actions/setup-python@v1
         with:
           python-version: '3.10'
-      - name: copy config
-        run: cp src/config.example.py src/config.py
       - name: install black
         run: pip3 install black==22.3.0
       - name: run black
@@ -115,6 +109,9 @@ jobs:
         run: docker-compose up --scale daemon=1 --build -d
       - name: run e2e tests
         run: docker run --net mquery_default -v $(readlink -f ./samples):/mnt/samples mquery_tests
+      - name: get run logs
+        if: always()
+        run: docker-compose logs
       - name: stop docker compose
         if: always()
         run: docker-compose down

.gitignore

@@ -1,6 +1,5 @@
 *.pyc
 venv/
-config.py
 .vscode
 .idea
 .mypy_cache

INSTALL.md

@@ -32,7 +32,6 @@ Docker compose dedicated for developers.
 ```
 git clone --recurse-submodules https://github.com/CERT-Polska/mquery.git
 cd mquery
-cp src/config.docker.py src/config.py
 # now set SAMPLES_DIR to a directory with your files, and INDEX_DIR to
 # empty directory for database files to live in. By default database will
 # expect files in ./samples directory, and keep index in ./index.

deploy/docker/daemon.Dockerfile

@@ -8,6 +8,5 @@ RUN ls /tmp/requirements*.txt | xargs -i,, pip --no-cache-dir install -r ,,
 
 COPY "src/" "/app"
 RUN chmod +x "/app/daemon.py"
-COPY "src/config.docker.py" "/app/config.py"
 
 ENTRYPOINT ["/app/daemon.py"]

deploy/docker/web.Dockerfile

@@ -17,5 +17,4 @@ RUN ls /tmp/requirements*.txt | xargs -i,, pip --no-cache-dir install -r ,,
 
 COPY "src/." "."
 COPY --from=build "/app/build" "./mqueryfront/build"
-COPY "src/config.docker.py" "config.py"
 CMD ["uvicorn", "app:app", "--host", "0.0.0.0", "--port", "5000"]

docker-compose.dev.yml

@@ -27,7 +27,8 @@ services:
       - "redis"
       - "ursadb"
     environment:
-      - "MQUERY_PLUGINS=${MQUERY_PLUGINS}"
+      - "REDIS_HOST=redis"
+      - "MQUERY_BACKEND=tcp://ursadb:9281"
   dev-daemon:
     build:
       context: .
@@ -42,7 +43,8 @@ services:
       - "redis"
       - "ursadb"
     environment:
-      - "MQUERY_PLUGINS=${MQUERY_PLUGINS}"
+      - "REDIS_HOST=redis"
+      - "MQUERY_BACKEND=tcp://ursadb:9281"
   ursadb:
     image: mqueryci/ursadb:v1.5.1
     ports:

docker-compose.e2etests-local.yml

@@ -25,6 +25,9 @@ services:
     depends_on:
       - "redis"
       - "ursadb"
+    environment:
+      - "REDIS_HOST=redis"
+      - "MQUERY_BACKEND=tcp://ursadb:9281"
   dev-daemon:
     build:
       context: .
@@ -38,6 +41,9 @@ services:
     depends_on:
       - "redis"
       - "ursadb"
+    environment:
+      - "REDIS_HOST=redis"
+      - "MQUERY_BACKEND=tcp://ursadb:9281"
   ursadb:
     image: mqueryci/ursadb:v1.5.1
     ports:

docker-compose.yml

@@ -15,7 +15,8 @@ services:
       - "redis"
       - "ursadb"
     environment:
-      - "MQUERY_PLUGINS=${MQUERY_PLUGINS}"
+      - "REDIS_HOST=redis"
+      - "MQUERY_BACKEND=tcp://ursadb:9281"
   daemon:
     restart: always
     build:
@@ -30,7 +31,8 @@ services:
       - "redis"
       - "ursadb"
     environment:
-      - "MQUERY_PLUGINS=${MQUERY_PLUGINS}"
+      - "REDIS_HOST=redis"
+      - "MQUERY_BACKEND=tcp://ursadb:9281"
   ursadb:
     restart: always
     image: mqueryci/ursadb:v1.5.1

docs/README.md

@@ -3,6 +3,7 @@
 ## User guide
 
 - [Installation](../INSTALL.md): Installation instruction.
+- [Configuration](./configuration.md): Additional configuration options.
 - [Components](./components.md): More detailed description of mquery components.
 - [Indexing](./indexing.md): Indexing files is one of the most important things in
     mquery. In simple cases it can be solved without leaving the web UI, but

docs/configuration.md

@@ -0,0 +1,135 @@
+# Configuration
+
+There are three different things you can configure within Mquery: core, plugins
+and ursadb. Unfortunately, all are configured differently.
+
+## Mquery core configuration
+
+Mquery is configured with [typed-config](https://github.com/bwindsor/typed-config). There are two ways to pass every configuration field - with
+a config file, or a environment variable. For example:
+
+```ini
+[redis]
+host=redis-server.example.com
+
+[mquery]
+backend=tcp://ursadb-server.example.com:9281
+plugins=plugins.archive:GzipPlugin
+```
+
+This is a simple INI configuration file that mquery understands. It should
+be saved in a file called `mquery.ini`. The file should be in one of the
+following locations (checked in that order):
+
+* Mquery's working directory (usually `src` folder in the cloned repository)
+* In the current user's xdg config directory: `~/.config/mquery/mquery.ini`
+* In the system config directory: `/etc/mquery/mquery.ini`
+
+Alternatively, you can use environment variables to configure mquery. All
+field names are mapped intuitively to environment variables by joining
+the ini section name with a key name - for example, to change redis host
+value use `REDIS_HOST`. Environment variables take precedence over values
+from the config file!
+
+Currently, supported configuration keys are:
+
+- `redis.host`: Hostname of a main redis server.
+- `redis.port`: Port of a main redis server.
+- `mquery.backend`: URL to a ursadb instance (for example,
+  `tcp://ursadb-server:9281`)
+- `mquery.plugins`: List of supported plugins, separated by commas (for
+  example `plugins.archive:GzipPlugin, plugins.custom:CustomPlugin`)
+
+## Mquery plugin configuration
+
+In contrast to the core configuration, plugins can be configured dynamically.
+Every worker registers its list of active plugins in the database, and it's
+possible to configure them using the web UI:
+
+![](./plugin-config.png)
+
+This configuration mechanism is used by the plugins shipped with Mquery.
+Despite this, it's optional, and plugin authors don't have to use it.
+Since plugins are arbitrary code, plugins can read their configuration from
+anywhere they want, including the environment, other config files, etc.
+
+It's also easy to use the same config file for Mquery and plugins - see
+[example_typed_config_plugin.py](../src/plugins/example_typed_config_plugin.py)
+file for an example.
+
+## UrsaDB configuration
+
+UrsaDB is not technically part of Mquery, but both systems work closely
+together and depend on each other for optimal performance.
+
+Mquery currently does not allow you to configure UrsaDB nicely.
+You have to do it "manually", by connecting with `ursacli` program to the
+TCP port exposed by UrsaDB. This program is built together with UrsaDB, and
+available in all official docker images. You can execute it in docker-compose
+like this:
+
+```
+sudo docker-compose -f docker-compose.dev.yml exec ursadb ursacli
+```
+
+Or you can download the latest ursadb release and run a client from there.
+
+To set a configuration field, issue a command like this:
+
+```
+$ ursacli
+ursadb> config set "database_workers" 10;
+```
+
+The configuration keys are already documented in the UrsaDB's docs here:
+https://cert-polska.github.io/ursadb/configuration.html. We won't copy
+all relevant information here, but the most important config keys are:
+
+* `database_workers` - How many tasks can be processed at once (you can
+  increase this for strong servers, restart the database to apply).
+* `merge_max_files` - Biggest supported dataset size. UrsaDB keeps indexed
+  files in so-called "datasets". The fewer datasets the faster the database is,
+  but you might not want overly huge datasets (at some point merging datasets
+  becomes very slow, and you get diminishing returns for merging them). Decide
+  on the value before indexing your files. Good values include the default
+  (infinite), 10 million, and 1 million.
+* `merge_max_datasets` - a very memory and CPU-intensive operation.
+  If your database OOMs during indexing, consider lowering this number and
+  the number of `database_workers` that are merging in parallel.
+
+## .env file
+
+Finally, in the main directory of the repository there is a file named `.env`.
+Mquery does not use it in any way, but it's read by Docker.
+
+```bash
+$ cat .env
+# This file is only relevant for docker-compose deployments.
+
+# Directory where your samples are stored. By default you have to copy them
+# to ./samples subdirectory in this repository.
+SAMPLES_DIR=./samples
+# Directory where the index files should be saved. By default ./index
+# subdirectory in this repository.
+INDEX_DIR=./index
+```
+
+If you use docker-compose to start mquery, you can use this file to specify
+a location on the host for your samples_dir and index_dir. These variables are
+then used when creating containers. See for example ursadb container spec:
+
+```yaml
+  ursadb:
+    restart: always
+    image: mqueryci/ursadb:v1.5.0
+    ports:
+    - "127.0.0.1:9281:9281"
+    volumes:
+    - "${SAMPLES_DIR}:/mnt/samples"
+    - "${INDEX_DIR}:/var/lib/ursadb"
+    user: "0:0"
+```
+
+As you can see, variables from `.env` are used to specify mount point for
+the data volumes. You can also ignore this file, and edit docker-compose
+directly to your liking.

docs/plugins.md

@@ -13,19 +13,19 @@ by plugins.
 
 ![](plugin-config.png)
 
-To add a new plugin to the system, you need to change PLUGINS key in
-`config.py` for bare metal setup. For example:
+To add a new plugin to the system, you need to change mquery.plugins key in
+[the config](./configuration.md). For example:
 
 ```python
-PLUGINS = ["plugins.mwdb_uploads:MalwarecageUploadsMetadata"]
+[mquery]
+plugins=plugins.mwdb_uploads:MalwarecageUploadsMetadata
 ```
 
 To load a plugin `MalwarecageUploadsMetadata` from `plugins.mwdb_uploads`
 module.
 
-To load plugins with docker-compose deployment, you can change
-`MQUERY_PLUGINS` environment variable in the container to load existing
-plugin, but to load your own plugin you need to create your own image.
+Remember that you can also use environment variable MQUERY_PLUGINS to do the
+same thing - this may be useful for docker-based deployments.
 
 ## Filter plugins
 

requirements.plain.txt

@@ -9,3 +9,4 @@ yara-python
 yaramod
 cachetools
 pyjwt[crypto]
+typed-config

requirements.txt

@@ -22,3 +22,4 @@ yara-python==4.1.3
 yaramod==3.12.1
 PyJWT[crypto]==2.3.0
 rq==1.11.1
+typed-config==1.3.2

src/.dockerignore

@@ -1,4 +1,3 @@
-config.py
 .pytest_cache
 .mypy_cache
 __pycache__

src/app.py

@@ -3,7 +3,7 @@
 from threading import Lock
 
 import uvicorn  # type: ignore
-import config
+from config import app_config
 from fastapi import (
     FastAPI,
     Body,
@@ -48,9 +48,9 @@
 )
 
 
-db = Database(config.REDIS_HOST, config.REDIS_PORT)
+db = Database(app_config.redis.host, app_config.redis.port)
 app = FastAPI()
-plugins = PluginManager(config.PLUGINS, db)
+plugins = PluginManager(app_config.mquery.plugins, db)
 plugin_lock = Lock()
 
 

src/config.py

@@ -0,0 +1,34 @@
+from typedconfig import Config, key, section, group_key  # type: ignore
+from typedconfig.source import EnvironmentConfigSource, IniFileConfigSource  # type: ignore
+import os
+
+
+@section("redis")
+class RedisConfig(Config):
+    host = key(cast=str, required=False, default="localhost")
+    port = key(cast=int, required=False, default=6379)
+
+
+@section("mquery")
+class MqueryConfig(Config):
+    backend = key(cast=str, required=False, default="tcp://127.0.0.1:9281")
+    plugins = key(cast=str, required=False, default="")
+
+
+class AppConfig(Config):
+    redis = group_key(RedisConfig)
+    mquery = group_key(MqueryConfig)
+
+
+def _config_sources():
+    return [
+        EnvironmentConfigSource(),
+        IniFileConfigSource("mquery.ini", must_exist=False),
+        IniFileConfigSource(
+            os.path.expanduser("~/.config/mquery/mquery.ini"), must_exist=False
+        ),
+        IniFileConfigSource("/etc/mquery/mquery.ini", must_exist=False),
+    ]
+
+
+app_config = AppConfig(sources=_config_sources())