mist-analytics v1.0.4

Mist

1. About Mist
2. Mist Set-Up
3. Mist Usage
4. FAQ
5. Future Wish-List + Contact Us

PART 1 - About Mist

Mist is an APM development tool for gathering metrics on unpublished CloudFlare workers (also known as serverless functions). With Mist, developers gain access to request and response data visualizations for their CloudFlare workers whilst running them in a local dev environment. Utilizing a smart system of interconnected local ports, the application "records" a session of requests which are then displayed across several visual mediums. This helps provide the developer with insight into the performance of their CloudFlare worker, opening the door for meaningful comparison and code optimization prior to deployment. Mist is currently in beta and under development.

PART 2 - Mist Set-Up

Option A - Using the npm package mist-analytics

1. Download NPM package here

Option B - Using the GitHub repository

1. Clone this repo into your local machine in the same directory where the worker you want to test is located. git clone https://github.com/oslabs-beta/mist.git
2. Install the dependencies npm install
3. Create a .env file in the root of the mist directory
   • Set WORKER_NAME equal to the exact name of your worker.
   • SetMY_URI equal to your postgresSQL database link (see next step).
4. Set up a postgreSQL database-- we recommend using elephantSQL-- and link it to the SQL schema
   • Copy the link to your empty database
   • Paste that link into the ENV file and save it as myURI
   • Open up the mist directory in your terminal
   • Run the following command: psql -d <url from elephantSQL> -f db_template.sql
5. Run the following scripts to start up the app:
   • npm run dev to start the GUI on localhost:8080
   • node server/server.js to start the server listening on your worker
   • node --require './server/tracing.js' server/miniflare to start your worker

PART 3 - mist Usage

1. Once all servers are running and your Worker is ready to be tested, navigate to localhost:8080. Here, to initiate the metric recording session click the Start button.
2. Next, navigate to localhost:8788 and fire off your Worker that is being tested. Allow for 5 seconds to elapse for data collection, prior to firing off the Worker again. This will ensure that the data is properly traced through Miniflare. Test whatever functionality you would like to see reflected in that session's metrics.
3. Once you have fired off all Worker invocations for that session, navigate back to localhost:8080 and click the Stop button to end the session.
4. In order to then display your session data, click the Generate Metrics button and the charts will display your metrics!
5. After you are finished analyzing this session data, click the Reset Metrics.

• Check out our Medium article for more information.
• Visit our website for documentation and further instructions on usage.

Logs

The logs that are retrieved from the Worker you are developing will be stored in an object called metrics which is composed of key that make up the logistics of the metrics information gathered. Method : stores the HTTP Request method/ URL : is set to the endpoint that the Worker you are testing is firing HTTP requests from. Status : the value is equal to whether the HTTP Request is successful 200 or a Bad Request 400 for example. Response Time : time set in milliseconds that is equal to the duration it took for the HTTP request to complete. Session Number : set to the current session number in your database. Start Time : set equal to the time it was when you clicked the start button on mist.

Request Time vs. Start Time

The data from the logs is plotted on this chart to show you the response time of the worker vs. the time it was fired off in the session you just ran.

Success vs. Error Rate

This pie chart offers a quick look at the rate of successes vs. errors on your worker from your previous session.

Past Sessions Average Response Time

When you fire off a worker, data from the last 5 sessions in which you tested that worker is aggregated and the response times are averaged. That data is rendered to this bar graph to give you a comprehensive comparison of how this worker has performed over sessions.

PART 4 - FAQ

Q: I ran a session but still am not seeing any data appear.

A: Make sure you hit the Generate Metrics button to render your data. If Mist is showing metrics from a previous session, make sure you hit Reset Metrics and then run a new session by pressing the Start button.

Q: I want to see a record of all logs from a specific worker. How can I access this data?

A: While this feature is still in development, you can access this data from your postgreSQL database by running the following command: SELECT * FROM "public"."metrics" WHERE worker = <your-worker-name>

Q: I fired off my Worker severeal times but I am not seeing all of that data reflected in the logs. What happened?

A: What most likely happened here is that the tracer did not have time to process your request to the Worker and store it in the database. Try waiting approximately 5 seconds between invocations of the Worker to ensure that all data is collected without interference.

PART 5 - Future Wish-List + Contact Us

Mist is an ever evolving open source product. If you’d like to contribute, please reach out to a member of the team with your proposal so we can incorporate you into the development process.

Future development goals

• Optimize OpenTelemetry controller by adding a batchprocessor in tracing.js. Current tracing operation takes around 3 seconds per HTTP request.
• Store user's worker file in a variable and run from the command line. Currently, worker is accessed in miniflare.js through path.resolve.
• More worker-specific metrics (CPU median time, subrequests, performance, etc.)
• Support for multiple workers running at once
• Modularize server, routing, controllers for Flare
• Update .env file with secrets

Team

• Anthony Marin @anthonymarinated | Linkedin
• Eric Lemay @StygianLiege | Linkedin
• Kristen Althoff @KristenWAlthoff | Linkedin
• Leo Rinke De Wit @lrinkedewit | Linkedin
• Stewart Elmore @selmore7 | Linkedin