1,028 reads

No More Heavy RAM Memory Consumption: Apply These 3 Secret Techniques

by AspectoMarch 29th, 2021

Too Long; Didn't Read

Docker Desktop on Mac allocates by default only 2GB of RAM to all containers in total. Each typescript service was using ~600MB of RAM out of the total 2GB available for all containers. The price to pay here is a longer compilation time. Instead of using tsc, it is possible to compile the code with tsc and monitor file changes with nodemon. If you want to run code in production that was already compiled and tested in the CI, you can only benefit from setting this option.

Company Mentioned

featured image - No More Heavy RAM Memory Consumption: Apply These 3 Secret Techniques

It turns out that running

ts-node-dev / ts-node

is constantly consuming hundreds of megabytes of RAM even for small and simple applications.

In development, it is usually not a big concern, however, it can be, if your application is running inside a docker container with limited resources (for example, with Docker Desktop on Mac which allocates by default only 2GB of RAM to all the containers in total).

Typescript code should be transpiled to Javascript which can be done either before running the process (

tsc

), or in runtime (

ts-node

The most efficient way is transpiling before running, however, this isn’t as developer-friendly since it takes forever.

ts-node-dev

loads everything into memory then watches the changes the developer is making and transpiles the project fast on every change.

We encountered the issue while building a demo application to showcase our product at Aspecto.

We were running multiple typescript services with docker-compose and started seeing arbitrary

ts-node-dev

processes exiting without even running the application, displaying the message “Done in 79.06s”.

This was due to a lack of memory. Each typescript service was using ~600MB of RAM out of the total 2GB available for all containers.

After digging a bit, we found a few possible solutions and wanted to share them.

Run

ts-node-dev

with option

--transpile-only

In our case, adding the

--transpile-only

option to

ts-node-dev

reduced the consumed RAM from ~600MB to ~170MB.

The price was that the typescript code would only be transpiled, and typechecking would be skipped. Most modern IDEs (vscode, web storm), has built-in typescript IntelliSense which highlights errors, so for us, it was a fair price to pay.

If you use

ts-node

to run code in production that was already successfully compiled and tested in the CI, you can only benefit from setting this option.

Compile the code with

tsc

and monitor file changes with

nodemon

Instead of using

ts-node-dev

, which consumes a lot of memory, it is possible to compile the application directly with

tsc

and then run it from dist/build like this:

node dist / index.js

For automatic reload on source file changes,

nodemon / node-dev

can be used.

This is our “start” script in package.json:

"scripts": {

  "start": "nodemon --watch src -e ts --exec \"(tsc && node dist/index.js) || exit 1\""

}

This approach reduced the RAM on our service from ~600MB to ~95MB (but there was still a spike in RAM to 600Mb for few seconds while

tsc

was compiling).

Unlink the previous option, this approach does check for typescript errors and warnings, and the service does not start if errors exist in the code.

The price to pay here is a longer compilation time. In our setup, it’s about 10 seconds from saving the file until the service restarts.

Increase Docker desktop available RAM

This is the easiest fix. Just allocate more Memory to Docker Desktop by going to Preferences => Resources => Memory, and increase the value.

While it fixes the immediate problem, the containers still consume a lot of memory, and if you have plenty of them, it might be a problem soon enough.

In addition, changing the default configuration should be done by every user that wants to run the system with docker-compose, which introduces complexity in installation and usage.