---
url: /advanced/debugging-and-performance-profiling.md
---
# Debugging & Performance Profiling

Are you seeing an error on an Open Library page?

The **first** thing you should do is check the applications streaming error logs:

```sh
docker compose logs --tail=10 -f web
```

Next, you can add `?debug=true` to the problematic url in order to reveal the stack trace or profile data.

This trick is especially useful when you see a nebulous error like:

```text
"Sorry. There seems to be a problem with what you were just looking at. We've noted the error 2019-02-05/193353339339 and will look into it as soon as possible. Head for home?"
```

If the request is a POST, use the browser to inspect the HTML form element and add `?debug=true` to the `action` url. If/when the request fails, you should see a useful stack trace.

## Using a Debugger

Using a debugger is one of the best ways to understand how code works and to find bug fixes. We use [VS Code](https://code.visualstudio.com/) (along with its [Python extension](https://marketplace.visualstudio.com/items?itemName=ms-python.python)) to remote debug into the docker container; we can't use regular debugging because the docker container behaves like a virtual machine.

![Debugger in action](https://user-images.githubusercontent.com/6251786/56706388-bd889e00-66e2-11e9-9d9b-449f0458305a.gif)

To use:

1. `docker compose up -d` as usual and login as admin.
   * *Note*: This has to be done before changing the compose file; see <https://github.com/internetarchive/openlibrary/issues/2122>.
2. Edit around line 6 of `compose.yaml` to say `workers 1` instead of `4` and `timeout` to be `180000000` instead of `180`
   * *Note*: By default, a worker only lives for 500 seconds -- this will interrupt the debugger; this setting should be increased (by e.g. 10x) or removed.
3. `docker compose up -d`
4. Go to <http://localhost:8080/admin/attach_debugger> and click "Start"
5. Go to the debug panel (Ctrl+Shift+D) in VS Code, and click "Python: Attach"
6. Debug!

The above instructions are for `web.py`. There's an issue for `fastapi` here: <https://github.com/internetarchive/openlibrary/issues/11829>

### Performance Profiling

Is a page taking a long time to load? Add `_profile=true` to the url in order to include a profile report. This will give you a report of both the front-end (client side) and the back-end (server side) load times. For a more granular breakdown of the client, follow the developer tools guide in your respective browser, e.g. <https://developers.google.com/web/tools/chrome-devtools/network-performance>.

### Website Performance

You can run a web page test via <https://www.webpagetest.org/> to obtain a better picture of openlibrary.org's performance on various devices given constraints.