Tracing

Observe APIs

Nowadays API Observability is already a part of every API development as it addresses many problems related to API consistency, reliability, and the ability to quickly iterate on new API features. When you design for full-stack observability, you get everything you need to find issues and catch breaking changes.

API observability can help every team in your organization:

Sales and growth teams to monitor your API usage, free trials, observe expansion opportunities and ensure that API serves the correct data.
Engineering teams to monitor and troubleshoot API issues.
Product teams to understand API usage and business value.
Security teams to detect and protect from API threats.

A central point for observation

We know that an API gateway offers a central control point for incoming traffic to a variety of destinations but it can also be a central point for observation as well since it is uniquely qualified to know about all the traffic moving between clients and our service networks.

The core of observability breaks down into three key areas: structured logs, metrics, and traces. Let’s break down each pillar of API observability and learn how with Apache APISIX Plugins we can simplify these tasks and provides a solution that you can use to better understand API usage.

Before enabling our plugins we need to install Apache APISIX, create a route, an upstream, and map the route to the upstream. You can simply follow getting started guide provided on the website.

Logs

Logs are also easy to instrument and trivial steps of API observability, they can be used to inspect API calls in real-time for debugging, auditing, and recording time-stamped events that happened over time. There are several logger plugins Apache APISIX provides such as:

And you can see the on the official website of Apache APISIX. Now for demo purposes, let’s choose a simple but mostly used http-logger plugin that is capable of sending API Log data requests to HTTP/HTTPS servers or sends as JSON objects to Monitoring tools. We can assume that a route and an upstream are created. You can learn how to set up them in the Getting started with Apache APISIX video tutorial. Also, you can find all command-line examples on the GitHub page

You can generate a mock HTTP server at mockbin.com to record and view the logs. Note that we also bind the route to an upstream (You can refer to this documentation to learn about more ).

The following is an example of how to enable the http-logger for a specific route.

note

To plugin settings, your can just put your mock server URI address like below:

{
  "uri": "http://mockbin.org/bin/5451b7cd-af27-41b8-8df1-282ffea13a61"
}

Once we get a successful response from APISIX server, we can send a request to this get endpoint to generate logs.


curl -i http://127.0.0.1:9080/get

Then if you click and navigate to the following our some recent logs are sent and we can see them:

Metrics are a numeric representation of data measured over intervals of time. You can also aggregate this data into daily or weekly frequency and run queries against a distributed system like . Or sometimes based on metrics you trigger alerts to take any action later. Once API metrics are collected, you can track them with metrics tracking tools such as Prometheus.

Apache APISIX API Gateway also offers to fetch your API metrics and expose them in Prometheus. Behind the scene, Apache APISIX downloads the Grafana dashboard meta, imports it to Grafana, and fetches real-time metrics from the Prometheus plugin.

Let’s enable prometheus-plugin for our route:

We fetch the metric data from the specified URL /apisix/prometheus/metrics.

curl -i http://127.0.0.1:9091/apisix/prometheus/metrics

You will get a response with Prometheus metrics something like below:

HTTP/1.1 200 OK
Server: openresty
Content-Type: text/plain; charset=utf-8
Transfer-Encoding: chunked
# HELP apisix_batch_process_entries batch process remaining entries
# TYPE apisix_batch_process_entries gauge
# HELP apisix_etcd_modify_indexes Etcd modify index for APISIX keys
# TYPE apisix_etcd_modify_indexes gauge
apisix_etcd_modify_indexes{key="consumers"} 17819
apisix_etcd_modify_indexes{key="global_rules"} 17832
apisix_etcd_modify_indexes{key="max_modify_index"} 20028
apisix_etcd_modify_indexes{key="prev_index"} 18963
apisix_etcd_modify_indexes{key="protos"} 0
...

And we can also check the status of our endpoint at the Prometheus dashboard by pointing to this URL http://localhost:9090/targets

As you can see, Apache APISIX exposed metrics endpoint is upon and running.

Now you can query metrics for apisix_http_status to see what HTTP requests are handled by API Gateway and what was the outcome.

In addition to this, you can view the Grafana dashboard running in your local instance. Go to http://localhost:3000/

You can also check two other plugins for metrics:

Datadog Plugin

Tracing

The third is tracing or distributed tracing allows you to understand the life of a request as it traverses your service network and allows you to answer questions like what service has this request touched and how much latency was introduced. Traces enable you to further explore which logs to look at for a particular session or related set of API calls.

Zipkin is an open-source distributed tracing system. is supported to collect tracing and report to Zipkin Collector based on Zipkin API specification.

Here’s an example to enable the zipkin plugin on the specified route:

We can test our example by simply running the following curl command:

curl -i http://127.0.0.1:9080/get

As you can see, there are some additional trace identifiers (like traceId, spanId, parentId) were appended to the headers:

"X-B3-Parentspanid": "61bd3f4046a800e7",
"X-B3-Sampled": "1",
"X-B3-Spanid": "855cd5465957f414",
"X-B3-Traceid": "e18985df47dab632d62083fd96626692",

Then you can use a browser to access http://127.0.0.1:9411/zipkin, see traces on the Web UI of Zipkin.

As you noticed, the recent traces were exposed in the above pictures.

You can also check two other plugins for tracing: