mpickering / graphql-engine-profiling-demo

graphql-engine Profiling Demo

NOTE: This repository contains important submodules. Either clone with the --recursive flag or do git submodule init && git submodule update after cloning.

---

A demo of profiling Hasura's graphql-engine using ghc-eventlog-socket and eventlog-live. Components:

• A graphql-engine executable that has been instrumented using ghc-eventlog-socket to stream its eventlog over a socket during execution.
• Another executable (in eventlog-influxdb) which reads data from the eventlog socket and inserts it into an InfluxDB database.
• A benchmarking tool (in ./bench) copied from the existing graphql-engine benchmark suite to generate traffic and make the profile more interesting.
• A Grafana instance to view the live eventlog data on a pretty webpage.
• A Docker Compose configuration to tie all of this together and run it all with a single command.

Usage

Build the images

NOTE: There are linux/amd64 and linux/arm64 images published on Docker Hub which are already listed in the docker-compose.yml, but they may be out of date or otherwise unacceptable, hence these instructions. Skip to Start the containers to use the prebuilt images.

Begin by building the base benchmarking image:

docker build ./bench/graphql-bench/app -t finleymcilwaine/graphql-bench:base

Then build the benchmarking image:

docker build ./bench -t finleymcilwaine/graphql-bench:main

Now build the base graphql-engine image:

docker build ./engine -f ./engine/base.Dockerfile -t finleymcilwaine/graphql-engine:base

Then build the graphql-engine image:

docker build ./engine -f ./engine/main.Dockerfile -t finleymcilwaine/graphql-engine:main

With the images built, the steps below should now work for you.

Start the containers

Run everything with just:

docker compose up

If it's your first time pulling the images, it may take a while (the graphql-engine image is about 1GB in compressed size).

The benchmark will just run on a loop. There will be a lot of docker compose log output from the the PostgreSQL and graphql-engine instances as they handle the requests from the benchmark.

Viewing the data

With Grafana

Navigate to localhost:3000 for the Grafana instance. Enter username admin and password admin.

The dashboards from ./grafana/dashboards are automatically loaded into the instance. To view and interact with them, select the "Dashboards" tab:

Query InfluxDB

To check the data in the InfluxDB instance directly, you can submit queries from your local command-line, e.g.:

curl -G 'localhost:8086/query?pretty=true&u=admin&p=admin' --data-urlencode "db=eventlog" --data-urlencode "q=SHOW MEASUREMENTS ON eventlog"

Outputs the available measurements.

curl -G 'localhost:8086/query?pretty=true&u=admin&p=admin' --data-urlencode "db=eventlog" --data-urlencode "q=SELECT * FROM \"gauge.eventlog.heap_size\""

Outputs the heap size measurement data.