Great video! I didn't know about Act, I'm gonna install it RN. I would like to add: I recommend using paths-ignore (inside push and below branches) to trigger a specific workflow only when a specific path is involved. For example if the docs folder is another project, if you push a readme update inside the project, all the workflow will run test again wasting you free minutes. With paths-ignore you can specify which path to ignore.
Act doesn't run 1:1 to GitHub Actions, doesn't support all third party actions, and is missing a few feature parity things that will crash with Act but works fine on Actions but it's still a huge life saver when testing 80% of things. Act also exposes act specific env variables so you can skip steps when running locally for example, which helps getting around some of these issues.
He already did a smaller video about it on the main channel 2 or 3 years ago. The only difference is that he used Node 12 back then which is either deprecated or downright removed now (can't remember what the Github blog post mentioned).
Hosted runners are life savers.. I basically use them for all my server work.. Once🎉 I push code, a self hosted runner running on the servers pulls the changes and restarts the server
I configured a self hosted runner on digitalocean. Do you know if it's possible to have multiple self hosted runners on the same VM? maybe with containers or something like that?
@@thekwoka4707 Like I said, it depends on how the runner is setup. The runners that Github itself uses are full VMs but you can setup up your own self-hosted runners that are just containers.
@Beyond Fireship Actually, if the project repo is public, then the workflows are free w/o limit (other than the 6 hour limit that all workflows have) the 2000 free minutes are for private repositories
I've been using this setup (actions/act) but with semantic-release for a while. It's perfection. Don't even have to think about bumping versions / changelogs / publishing. Also you can create a little workflow badge to put on the repo or to keep a link to. Nice for having an overview of the state of several repos.
Within the first 16 seconds, I had to laugh/smile multiple times. Your sense of humor is great. Also, the time and deep thoughts you spend int this work is clearly visible.
It depends. For me, I have a M1 laptop and it wasn't functioning properly when I set it up as a runner so I had to go about configuring a docker container as a runner so I could then run nested containers inside of it, which was excruciating.
As long as you recognise the warning they include with the option, if someone makes a pr on your public repo and it triggers some test, then whatever code they submitted will be run on your machine which may include malware.
I like to use Volta in my projects, handily you can then use Volta in your GHA pipeline. No need to specific node or npm/pnpm/yarn versions any more. It will use exactly the version the project was developed on. Which is fantastic when upgrading node version.
I can't tell you how much time I've wasted in the past, pushing 1 byte changes to make a CI pipeline run while debugging the flow, all the time wishing I could just run the damn yaml/json/whatever file locally to simulate the CI/CD server ... thanks for the "act" tip!
I would have liked a video like this when I first learned GitHub actions, but still now I have to configure a new more complex project so that's still a good timing for me❤
7:58 a use case for self-hosted runners is if you run docker containers on a vps and want to automate the creation of images without having to git pull and run the commands yourself every time
I use a slightly different approach where I run a container inside GitHub Actions and run all of the commands against that container which acts as a development container too. So I achieve a couple of things at the same time. I do not need act to test my actions because all of the logic behind my actions is containerized anyway.
8.5k views, 685 likes but only 25 comments (as of the time writing this) I really appreciate your videos and content, so here, take this free comment to sacrifice to the algorithm gods.
I open these videos out of routine, sometimes I don't even watch, they just play in the background. I like his voice. Sometimes I learn something, but usually I don't.
Act is pretty limited though. Doesn't work for trigger inputs and in my case, I had to actually run a nested container for my tests so Act was pretty much useless.
I've been doing some GitHub actions with install-nix-action and then running 'nix build' which covers all dependencies and declaratively builds the project. Maybe you can make a video on nix next.
Great video, Jeff! 🔥 Was so great to learn about act - I looked for a solution like this a while back and couldn't find anything. Quick question though: Is there any specific reason for using v2 over v3 of actions/checkout and action/setup-node?
Versioning with commit messages is a bad idea - you will run into conflicts during merges. Use git tags to version and inject tag values into your project files during build if you need to access the version from code. Git tags can be also used to trigger special actions which do not get triggered otherwise (e.g. run deploy, build docs, run database migrations…)
Assuming you're using semantic versioning and conventional commits in your project, you can use a tool like semantic release to automagically determine the next version number.
Have you had any experience shortening that feedback loop earlier on the process, let's say at the pre-commit timeframe? I ask because we would commit things up to the repo to fire off a pipeline and then get fredback a few minutes later. It felt nicer (to me) to abort our commit if our tests, or linting failed. Yes there is always the option to force it through, but I liked that immediate feedback locally before something small and dumb was sent up.
Yeah, unit tests and linting and formsttkng should be pre-commit since they should be very fast. Then they can also be in the PR and the PR can also have the longer processes like e2e tests, builds, etc.
I disagree with npm CI for these. I use pnpm and cache the store (and even node modules itself) to speed up actions (cache key is the package-lock hash) Npm is slow, and CI prevents any caching options.
Where was written the playwright test code? I only saw you linking it to the actions or something like that, but you never made the code that makes the test. I don't get it.
I have used github actions to trigger tests on every push. But I experienced very slow behavior. Even for a very small small project it took tens of minutes to complete. Is it normal to take so long?
Hey) The video is awesome))) Thank you. I have a question thought: In a github issue page there s a button "create branch" (inside the Development section of the right panel). Is there a way to trigger an action that changes a state of the corresponding issue to inProgress when a branch is created using this button? I use github project to manage issues. Its awesome too)
I wonder whether anyone has setup completely standardized projects for use cases like "Python3 in a Docker image", including GitHub actions, linting and testing and whatnot.
Great video! I didn't know about Act, I'm gonna install it RN.
I would like to add:
I recommend using paths-ignore (inside push and below branches) to trigger a specific workflow only when a specific path is involved.
For example if the docs folder is another project, if you push a readme update inside the project, all the workflow will run test again wasting you free minutes. With paths-ignore you can specify which path to ignore.
Yes, I found this trick recently and this was great for only building/testing those modules that changed.
Act doesn't run 1:1 to GitHub Actions, doesn't support all third party actions, and is missing a few feature parity things that will crash with Act but works fine on Actions but it's still a huge life saver when testing 80% of things.
Act also exposes act specific env variables so you can skip steps when running locally for example, which helps getting around some of these issues.
@@marty0678 I agree, something may fail in act but work in actions.
I tried it several times but it seems like they don't work on macOS? I might be missing sth but from my experience act didn't work for me.
@@sungjuyea4627 I use it on macOS
This is my favourite type of content, covering stuff nearly every developer uses or should be using and showing tips, tricks etc.
The timing of this tutorial... I've been meaning to use GitHub actions to maintain a npm package... This couldn't have come out at a better time
He already did a smaller video about it on the main channel 2 or 3 years ago. The only difference is that he used Node 12 back then which is either deprecated or downright removed now (can't remember what the Github blog post mentioned).
What’s your package about ?
@@YuriG03042 I seem to have missed that video somehow 🙈
@@ウタ-u1g it's a fork of unmaintained calendar library for vue which I am slowly modifying for my office's needs
I've just used it to add latest blogs to my readme lol
Hosted runners are life savers.. I basically use them for all my server work.. Once🎉 I push code, a self hosted runner running on the servers pulls the changes and restarts the server
I configured a self hosted runner on digitalocean. Do you know if it's possible to have multiple self hosted runners on the same VM? maybe with containers or something like that?
@@woops30 no idea
@@woops30 it is, simply start multiple and make sure they are named differently
Currently working on setting CI/CD on a project. what a time to upload the video👌
Github Actions actually spins up a whole vm. There are very specific circumstances in which this is important as I spent two weeks learning.
What circumstances?
TIL in theory, you can self-hosted github actions if there's a way to redirect the builder
To be fair, that would mean you are not using github lmao
It doesn't have to use a whole vm. I configured a docker container to act as a runner so I could run nested docker containers inside of it.
@@Bash70 it still uses a full VM to run your action container.
@@thekwoka4707 Like I said, it depends on how the runner is setup. The runners that Github itself uses are full VMs but you can setup up your own self-hosted runners that are just containers.
@Beyond Fireship Actually, if the project repo is public, then the workflows are free w/o limit (other than the 6 hour limit that all workflows have)
the 2000 free minutes are for private repositories
I had no idea something like act existed, you just saved me a shit ton of time!
I've been using this setup (actions/act) but with semantic-release for a while. It's perfection. Don't even have to think about bumping versions / changelogs / publishing. Also you can create a little workflow badge to put on the repo or to keep a link to. Nice for having an overview of the state of several repos.
Within the first 16 seconds, I had to laugh/smile multiple times. Your sense of humor is great. Also, the time and deep thoughts you spend int this work is clearly visible.
Calling it life-saver is an understatement to be honest.
The selfhosted runners are super easy to set up, i did it and that way you also dont hsve any rate limits!
It depends. For me, I have a M1 laptop and it wasn't functioning properly when I set it up as a runner so I had to go about configuring a docker container as a runner so I could then run nested containers inside of it, which was excruciating.
As long as you recognise the warning they include with the option, if someone makes a pr on your public repo and it triggers some test, then whatever code they submitted will be run on your machine which may include malware.
@@Bash70 oh yeah I hadn't considered Macs, I ran mine on Linux but yeah I guess other OSs could definitely make things more difficult
Imagine watching a entire course with this intonation video after video for hours.
I had act but never knew how to use it right! Thanks for showing us!
I like that fact he is smart and advertises his products.
I like to use Volta in my projects, handily you can then use Volta in your GHA pipeline. No need to specific node or npm/pnpm/yarn versions any more. It will use exactly the version the project was developed on. Which is fantastic when upgrading node version.
Great video! Taught me enough to get automation on.
Love it! Straight and concise! Thank you!
I can't tell you how much time I've wasted in the past, pushing 1 byte changes to make a CI pipeline run while debugging the flow, all the time wishing I could just run the damn yaml/json/whatever file locally to simulate the CI/CD server ... thanks for the "act" tip!
I knew about GH actions but I never knew about act! Thanks
I would have liked a video like this when I first learned GitHub actions, but still now I have to configure a new more complex project so that's still a good timing for me❤
Thanks for sharing! I started looking a bit more into GitHub actions, yesterday
I did not know about the act package, that's a pretty nifty tool
Cool, I understood some of the words in this one!
Master branch keep the faith brother
I didn't know about `act`. That's cool.
The github YAML files look beautiful, very intuitive.
7:58 a use case for self-hosted runners is if you run docker containers on a vps and want to automate the creation of images without having to git pull and run the commands yourself every time
Top-notch tutorial, as always 👏
Great video - lots of useful resources and content
6:13
I fucking love cron magic strings, this for sure won't backfire in the future.
For an 'automated' solution that sure sounds complex and convoluted.
I will debug myself to learn more from you
Best video, so far !!!
I use a slightly different approach where I run a container inside GitHub Actions and run all of the commands against that container which acts as a development container too. So I achieve a couple of things at the same time. I do not need act to test my actions because all of the logic behind my actions is containerized anyway.
8.5k views, 685 likes but only 25 comments (as of the time writing this)
I really appreciate your videos and content, so here, take this free comment to sacrifice to the algorithm gods.
I open these videos out of routine, sometimes I don't even watch, they just play in the background. I like his voice. Sometimes I learn something, but usually I don't.
I just love your content. ❤
Love your videos bro. Thanks.
Super awesome 😎 and informative. Thanks a lot!
Thanks. Didn't know Act before, so every time I want to change the workflow I have to wait for the whole run like an idiot.
Act is pretty limited though. Doesn't work for trigger inputs and in my case, I had to actually run a nested container for my tests so Act was pretty much useless.
watched it now, got the basic idea but im currently not using or planning to use or learn it now. I'll be back soon :)
Thank you for nektos/act.
svelte is gold but react is gem!
I've been doing some GitHub actions with install-nix-action and then running 'nix build' which covers all dependencies and declaratively builds the project. Maybe you can make a video on nix next.
0:06 is my life motto
This is truly great but I just wish there was a feature that would automate that one thing I need to manually do every day at my current project
Great video, Jeff! 🔥 Was so great to learn about act - I looked for a solution like this a while back and couldn't find anything. Quick question though: Is there any specific reason for using v2 over v3 of actions/checkout and action/setup-node?
Yuh much agreed. @v2 is going to give a deprecation warning for node 12.
you mean you don’t have like dozen of commits “CI TEST”? would love a breakdown on that act tool!
right when i needed it!
Do GitLab next? Would help a lot of us which have Corporate GitLab servers
All these features are not compatible between the different services. That’s why I try to stick to common functionality.
ooh damn, thats very cool!
Beyond programming
I can't believe I didn't know about act
Thank you so much for this
How do you automate versioning? Do you have to version bump manually in yml with each release?
Use semantic versioning, it looks at your conventional commits to determine version based off the commit message
Versioning with commit messages is a bad idea - you will run into conflicts during merges. Use git tags to version and inject tag values into your project files during build if you need to access the version from code.
Git tags can be also used to trigger special actions which do not get triggered otherwise (e.g. run deploy, build docs, run database migrations…)
Assuming you're using semantic versioning and conventional commits in your project, you can use a tool like semantic release to automagically determine the next version number.
10x is not a verb, sir
*10xed
Did you “google” that?
Somebody is trying to 10x their grammar. And failing ...
10x ur mom
Not a Grammer class😢
Have you had any experience shortening that feedback loop earlier on the process, let's say at the pre-commit timeframe? I ask because we would commit things up to the repo to fire off a pipeline and then get fredback a few minutes later. It felt nicer (to me) to abort our commit if our tests, or linting failed. Yes there is always the option to force it through, but I liked that immediate feedback locally before something small and dumb was sent up.
Yeah, unit tests and linting and formsttkng should be pre-commit since they should be very fast.
Then they can also be in the PR and the PR can also have the longer processes like e2e tests, builds, etc.
Wright timing. ✅
0:00 that sound make subwoofers hit me in the back. aa
hi I noticed that you use v2 actions that produce tons of warnings about nodejs deprecating, can you use v3?
I disagree with npm CI for these.
I use pnpm and cache the store (and even node modules itself) to speed up actions (cache key is the package-lock hash)
Npm is slow, and CI prevents any caching options.
Is it possible to configure github actions to auto deploy to a custom server instead of big company servers like azure or aws?
Ansible?
If you can do it from the command line, yes
very wise indeed
Please make a video about auto merging
what I don t like about Github Actions is that it doesn t support yml aliases which makes the file quite cluttered for big ci cds pipelines
You have both main and master branches?
What is this madness branching strategy? :))
tx! awesome video..
Where was written the playwright test code? I only saw you linking it to the actions or something like that, but you never made the code that makes the test. I don't get it.
Hey fireship do tutorials on legend state for react
I have used github actions to trigger tests on every push. But I experienced very slow behavior. Even for a very small small project it took tens of minutes to complete. Is it normal to take so long?
Hey) The video is awesome))) Thank you.
I have a question thought: In a github issue page there s a button "create branch" (inside the Development section of the right panel). Is there a way to trigger an action that changes a state of the corresponding issue to inProgress when a branch is created using this button?
I use github project to manage issues. Its awesome too)
Could you do some dewops stack?
Brb gonna write-up a guide on Github Actions, am not gonna plow through life without a CI/CD engine
I wonder how firebase functions are work with GitHub actions?
If I give gpt the link of the video can build the app automatically?
Please make a video about nx
I don't think it's technically true that GitHub spins up a container when you start a workflow. I believe they connect you to a full VM.
They use microvms and spin up containers inside them for certain action types. But a lot of it can be done using the base VM.
They're all containers!
I think I heard the basement slaves at 7:32 doing animations for the daily code reports
Please make a video about web componend
is act stable now? I had so many bad experiences with it in the past :(
Act is not working on M1 macbook.
Here’s a nickel, kid. Get yourself a real computer.
something important missing, you should have explain how to deploy the code on your production machine upon code push.
I gave up on making some ai that reads the code / commit messages and finds relevant memes to it and adds it to the or description
CV/Resume
Skills:
Ci/CD expert
Here are 27 hoops you can jump through to make your life easier
Could you please do a video on free tools and services? It would be very handy after the death of heroku
Dude if I implement the automated test while a release on git i will lose my job 😶… and before ai, i will be the one to make my job absolute 😶
Dude, I need your jokes workflow!!!
Im starting with CI/CD
Soo many things too install, it's overhwelming
Github action ftw
It’s like Microsoft Visual Studio has come to Linux!
The only thing missing here is caching the playwright installation
Ah yes, who doesn't love the model: "The first one is free..."
It's drone CI all the way for me.
Man, this AI voice is good!
idk i just use the google action lol aka release please
Wait, this is about testing my project, I’m out 😂
I wonder whether anyone has setup completely standardized projects for use cases like "Python3 in a Docker image", including GitHub actions, linting and testing and whatnot.
Python in docker, what a nightmare!
VS code Theme and icon Please
Use ASDF instead of node setup