Optimize docker memory #976
Conversation
📦 A new release has been made for this pull request.To play around with this PR, pull Images are available for x86_64 and ARM64.
|
ninele7
force-pushed
the
optimize-docker-memory
branch
from
9940687 to
1666045
Compare
ninele7
force-pushed
the
optimize-docker-memory
branch
from
1666045 to
a325308
Compare
ninele7
force-pushed
the
optimize-docker-memory
branch
from
ce3d97b to
1e85acb
Compare
Dockerfile
Outdated
| @@ -34,4 +45,6 @@ ENV NODE_ENV production | |||
| ENV COMMIT_HASH $COMMIT_HASH | |||
| ENV BUILD_DATE $BUILD_DATE | |||
|
|
|||
| CMD ["tini", "--", "yarn", "start"] | |||
| RUN echo "DATABASE_URL=$(node dist/scripts/print-database-url.js)" > .env | |||
There was a problem hiding this comment.
we shouldn't bake the DATABASE_URL into the image, as it potentially exposes a secret, won't work for the images published publicly on Docker Hub, and makes it harder to change
I think it should work without this step?
There was a problem hiding this comment.
Now it can't expose any secret because location of file with SQLite DB isn't a secret.
But I changed it to be more configurable.
| @@ -14,18 +14,29 @@ COPY package.json . | |||
| COPY yarn.lock . | |||
|
|
|||
| RUN yarn install --prod | |||
| RUN cp -R node_modules /usr/app/prod_node_modules | |||
There was a problem hiding this comment.
I would recommend doing yarn install (all dependencies) in this dependencies layer, then not copying node_modules and instead running yarn install --prod in the runner image to simplify it a little
There was a problem hiding this comment.
I've done it to avoid downloading prod dependencies twice. That way it seems to work faster. But if you prefer calling downloading twice I will change it.
|
🚀 Released in Release v2.4.4. |
When running this bot in docker container on my VPS server I encountered high ram usage. After investigation I found out that it runs with chain of yarn/npm/tsx/run-with-database-url.ts.
In this PR I:
tsc.tsc.run-with-database-url.tswith simpler solution that doesn't run additional node processes in runtime.yarn startwith direct invocation of node indockerfile.Before PR docker stats displayed 300-400mb of memory usage. Now it's 60-120mb.