Elastic

How to integrate Elasticsearch with OpenReplay and see backend errors alongside session recordings.

If you are using Kibana dashboard:

1. Login to your Kibana dashboard.
2. Go to Dev Tools > Console.
3. Copy the following console code to your console and run it.

If you are using CURL commands:

1. Execute the following cURL command in your terminal.

Note: by default, this integration will search for logs inside any index that matches *log*, if you have a specific index for logs that doesn’t match this pattern, please change the name of the index in line 12 in the following command. If you want to specify more than 1 index, you can separate the names with ,.

POST /_security/api_key
{
  "name": "openreplay-api-key",
  "role_descriptors": {
    "openreplay-role": {
      "cluster": [
        "all"
      ],
      "index": [
        {
          "names": [
            "*log*"
          ],
          "privileges": [
            "read"
          ]
        }
      ]
    }
  }
}

If you used any of the previous methods to generate an API key, you will get a result as the following:

{
  "id" : "eQWAIG0Bo0VqB8HXFH9-",
  "name" : "openreplay-api-key",
  "api_key" : "dZ5ycVRJTU-5UW_RYfi1_w"
}

Make sure to copy the id and the api_key as we need them for our integration. For more information about creating an API key, please refer to this documentation.

Put your host address, port, id and api_key in OpenReplay dashboard under ‘Preferences > Integration’. If you changed the index when generating the api_key, please specify the name in indexes.

To link a Elasticsearch event with the recorded user session, a unique token has to be propagated from your frontend to your backend on each request you want to track. This can be done using a custom HTTP header. In the below example, we use the fetch function to send that header.

const headers = {
  Accept: 'application/json',
  'Content-Type': 'application/json',
};
if (tracker.getSessionToken()) { // or window.OpenReplay instead of tracker if you're using the snippet
  headers['X-OpenReplay-SessionToken'] = tracker.getSessionToken(); // Inject openReplaySessionToken
}
fetch('www.your-backend.com', {
  'GET',
  headers,
});

In order for OpenReplay to associate a Elasticsearch log entry with the recorded user session, a unique token has to be propagated as part of each backend error you wish to track.

Below is an example in Python using Monkey Patching.

import sys
import traceback
#...
old_tb = traceback.print_exception
old_f = sys.stdout
old_e = sys.stderr
OPENREPLAY_SESSION_TOKEN = None

class F:
    def write(self, x):
        if OPENREPLAY_SESSION_TOKEN is not None and x != '\n':
            old_f.write(f"[openReplaySessionToken={OPENREPLAY_SESSION_TOKEN}] {x}")
        else:
            old_f.write(x)

    def flush(self):
        pass

def tb_print_exception(etype, value, tb, limit=None, file=None, chain=True):
    if OPENREPLAY_SESSION_TOKEN is not None:
        value = type(value)(f"[openReplaySessionToken={OPENREPLAY_SESSION_TOKEN}] " + str(value))
    old_tb(etype, value, tb, limit, file, chain)

traceback.print_exception = tb_print_exception

sys.stderr = F()

The name of the tag openReplaySessionToken is case sensitive.

By default, we look for each log associated to an OpenReplay session using the attributes message and utc_time, we are only interested in error logs, to identify error logs we check the attribute tags to contain the value error.

If you encounter any issues, connect to our Slack or check out our Forum and get help from our community.