3

I'm using Jenkins Multibranch Pipeline to handle my CI/CD and I'd like to know if is there a way to cache my node_modules folder?

Every time I push a branch, I execute yarn to install the dependencies. when is master branch, it took 1s to execute this step because it's already cached. Otherwise, when is a new branch, yarn took 2 minutes because is a new environment. Is there a way to cache node_modules globally to be used by all branches/environments?

Pedro Arantes
  • 133
  • 1
  • 1
  • 5

3 Answers3

4

It is possible to do this (not with any built-in steps; you essentially have to either use a global cache or write your own caching tool), but I do not recommend it. When I've tried this in the past, I often ran into race conditions. For example, one build would write to the cache while another build simultaneously reads from the cache, or two builds would write to the cache simultaneously, etc. I got a lot of builds with corrupted modules as a result.

Caches do not have these kinds of race conditions when limited to a single job (or in this case, a single branch for a multibranch job) because Jenkins can be configured not to run multiple builds of one job simultaneously. If you take extra steps to ensure that all of the jobs which might access the cache can't run simultaneously, then you could probably get away with a shared cache. But at that point, you're likely to lose the added speed benefit of sharing the cache anyway, since you've massively reduced Jenkins' capacity to run jobs in parallel.

The other scenario where you could get away with a shared cache is if you write to the cache extremely infrequently - and I really do mean extremely. If the cache is rarely written to, then there simply aren't many opportunities for these kinds of race conditions.

jayhendren
  • 2,952
  • 7
  • 15
  • Great explanation! I never could imagine the problem about race conditions but makes a lot of sense. I'll think another way to handle the installation time instead of creating a cache folder. – Pedro Arantes Jun 08 '19 at 22:39
2

Adjust Jenkins job "Build Environment" enter image description here see https://stackoverflow.com/questions/39829229/how-to-use-exclude-folder-option-in-delete-workspace-before-build-starts-secti/41728663

also add build step with shell/windows script with

npm install -D
Sasha Bond
  • 121
  • 2
1

The best way I have found so far is to use Docker containers with a docker file that does the installation of node_modules before running other steps.

I make sure all jobs of this multi branch pipeline run on the same node and keep the latest docker image for EACH branch on this node locally. That way unless the dependencies in packages.json change, then docker will use the cache to build the new image (or just generate the artifact you want to create and transfer it to the local workspace). If all branches have the same dependencies, they will all use the same cached image, if not they will just create a new one.

Example:

## CHOSE YOUR LOCAL BASE IMAGE
FROM debian:10.0

ADD YOUR DEPENDENCIES


ADD packages.json build-deps.sh


INSTALL DEPENDENCIES (we use a custom script)


RUN build-deps.sh


ADD THE REST OF YOUR CODE


This will break the cache assuming your code changes each time,


but previous docker build commands will be cached


(if your dependencies didn't change)


ADD my-code.tar.gz .


RUN YOUR COMMANDS


RUN run-your-scripts.sh

After building you image with docker build, you can then do soemthing like this to extract the artifact

docker run --name SOME-NAME YOUR-REPO:YOUR-TAG
sudo docker cp SOME-NAME:/path/to/artifact/artifact.tar.gz ${WORKSPACE}

I hope this helps

Pedro Graber
  • 111
  • 2