How to deploy your Node.js service(blog.gosquadron.com)
blog.gosquadron.com
How to deploy your Node.js service
http://blog.gosquadron.com/deploy-node
14 comments
I'm using Flightplan (https://github.com/pstadler/flightplan) which was posted on HN a while ago. All it does is provides a thin layer (albeit one with liberally applied DSL) on top of SSH which helps manage production/dev automated deployments.
+1 for using Node to provision Node apps. I have messed with flightplan and think it is awesome. It inspired me to make something that is a bit more like puppet (plugin based packaging, with helpers for common provisioning tasks) but in node (https://github.com/wesleytodd/wombat).
This is after two tries to get my company over to a Vagrant based setup, once using puppet and a second using bash scripts. Unfortunately neither got us off the ground fast enough to make it viable.
This is after two tries to get my company over to a Vagrant based setup, once using puppet and a second using bash scripts. Unfortunately neither got us off the ground fast enough to make it viable.
I would add the existing package managers to that (eg: yum+rpm or apt+dpkg).
Those + some bash scripts to sort out credentials and relationships between services - sorted!
Those + some bash scripts to sort out credentials and relationships between services - sorted!
This has built-in tests and automatic rollback of failed deployments among other cool features. I've built custom deployment systems before (in bash and make) and they're hard to scale from yourself to more people. Bugs in your deployment tools suck.
PSA: One of the creators of Squadron.
PSA: One of the creators of Squadron.
or setup a git server and a webbook that calls the bash file on push. The bash script would copy the files, npm install, and cycle pm2 or forever. This is how I set up my servers. Seems less complicated.
"its easy" followed by 10 pages on how to configure it.
really though, how is this any better than ansible?
really though, how is this any better than ansible?
Came here to make the same exact comment, ask the same question.
yes hoping OP will explain; so far I don't see the win.
I have a git server (gitolite) with a repo and four branches: main, development, testing and production.
When im ready to run something on testing, for example, I checkout testing, merge from development and push.
In the remote server, the code it the repo is running with pm2. I use a post-receive hook to npm update everything and pm2 restart.
If everything is ok, and I want to deploy into production, I do the same thing but merging production branch instead.
When I see tools like this, I tend to think that im missing something, or that there might be something wrong with what I do.
What do you think?
When im ready to run something on testing, for example, I checkout testing, merge from development and push.
In the remote server, the code it the repo is running with pm2. I use a post-receive hook to npm update everything and pm2 restart.
If everything is ok, and I want to deploy into production, I do the same thing but merging production branch instead.
When I see tools like this, I tend to think that im missing something, or that there might be something wrong with what I do.
What do you think?
The problem I have with setting up these things yourself is that I usually forget about how I set it up in the first place. Sure you might remember how you set it up 3 months down the road, but what about 3 years?
I try to use services or tools that do as much as possible out of the box. It means that I didn't have to manually configure much, which means it will be easier to understand after long periods of time. I know it isn't very hacker-like, but I personally feel there is a gray area that might create more problems than it solves, and I tend to stay on the "safe" side of things.
I try to use services or tools that do as much as possible out of the box. It means that I didn't have to manually configure much, which means it will be easier to understand after long periods of time. I know it isn't very hacker-like, but I personally feel there is a gray area that might create more problems than it solves, and I tend to stay on the "safe" side of things.
Alternatively, this is the kind of scenario I wrote Git-deliver for : https://github.com/arnoo/git-deliver
I'll take rsync over a 234234 line config / external dpeendencies and a tutorial longer than most node.js tutorials themselves.
Thanks for taking the time though.
Thanks for taking the time though.
Why use a python script?
Automatic deployment is usually 3-10 steps that need to be run in a remote server. How's this better that a simple script that sshs into the server/servers, CDs into the project dir and does `make update`, which pulls some branch, runs any migrations and does an `apache gracefully` or similar whatever needed for the update?