Hacker News Reader

CI has essentially become a commodity with a need for reliability.

For us, we moved from Travis to Github because Travis just wasn't reliable enough (and had terrible security). We'd be blocked from deploying for hours at a time because Travis had a random outage. I know it's way more complex than "boxes that run some stuff", but CI/CD isn't exactly rocket science.

We also noticed that by using a dedicated CI provider, we doubled our deployment downtime. We'd be blocked from deploying when either Travis or Github had issues. Whereas, we're only blocked now when Github has issues.

100% can back the claim about Travis. It is awful and very buggy.

For our case, we are migrating to the Google Cloud Build. More stable and features but also some things are still lacking. Nonetheless, a big step up from Travis.

I wish it was a commodity, but here I am forced to use Azure CI

We used to pay CircleCI $200/month to run CI on a widely-used open-source project, which was set up before GHA was available. Around one year ago we switched to GHA, which was significantly better, and cost $0.

At around the same time we made the switch, we got an opaque email from CircleCI about their new `performance plan` which would have increased the cost to $300/mo. This was more metered pricing.

Anyway GHA has been a breath of fresh air and I have convinced several open-source maintainers do dump CircleCI in favor of GHA.

When I read this, not judging or anything, no ulterior motive, I'm reminded Microsoft and embrace, extend, extinguish.

When and if they get rid of the competition, it likely won't be free aka 0 monetary costs again.

We also recently left them for self-hosted GHA runners. The other problem is that their pricing is insane - we're building exactly the same code for ~20% of the cost.

I could never understand their bills. I seemed to get dozens of small invoices throughout the month. Their whole concept of credits was completely puzzling too. Left to GHA.

I'd love for self-hosted GitHub Actions runners to work in ECS Fargate.

Agreed, the pricing is all wrong. That's the #1 reason I know people have actively moved off. Not competitive esp as all source control services have it included now.

> I genuinely don't understand how CircleCI is still a business.

In slight defense of Circle CI I think that could be applied to at least 50% of tech companies. Though CCI is perhaps more guilty than most of this transgression.

A lot of places probably set up CircleCI and don't want to invest into switching since the value add between services is so minimal.

Why start a new project with CircleCI? That I don't have a clue.

When GHA has decent instance sizes I imagine they'll eat a lot of CCIs lunches. The automatic test splitting is also a god send for parallelising builds.

Github Actions now has up-to 64-core Linux and Windows instances.

Thanks for pointing that out. Are these large runners somewhat new? I feel like I check every few months and I hadn't seen them before.

https://docs.github.com/en/actions/using-github-hosted-runne...

4 core | 16gb RAM | 150gb SSD is the next size up which is just about what I need.

Edit: September 1 was the announcement, so I was probably due to check back in: https://github.blog/changelog/2022-09-01-github-actions-larg...

This was my exact reaction to. I was a customer for CircleCI for years and it was my go to choice and introduced it to countless companies, they were clearly at first an engineering led company that obsessed about functional programming.

At some point after their series C they became rent-seeking, their support became increasingly distant, and the quality overall started degrading. Then I get an email that we need to pay for seats.. on top of paying for CI minutes.

It was an extra > $500 a month for no additional service, no additional usage, no nothing, and they would not budge on their new pricing plan. So I just moved all the stuff to GitHub who was barely mature enough at that point and never looked back.

Putting the finance department in charge of product decisions is a mistake with long term reproductions and CircleCI is only seeing the results of this. Treating your oldest customers like a financial liability is a mistake.

I agree this is really sad and I also hope everyone affected finds a new opportunity quickly. CI is a commodity so it's hard for CircleCI to sustain as a stand-alone business. I found DroneCI to be simplest and fastest CI tool I have used. In fact just recently Harness published some bold claims of being 4x faster than leading CI vendors(https://harness.io/blog/fastest-ci-tool). I have been using it myself and it's faster. The benefit of having CI, CD, Security testing and governance features makes it a compelling platform offering similar to GitHub & Gitlab.

I don't know how bad CircleCI is, but if it is worse than GH Actions, then that's quite an achievement ;) GH Actions kinda works, but I don't know who it has been designed for (especially compared to much more mature solutions like Gitlab CI). On top of that, GH Actions VMs seem to be extremely performance-throttled, and it can take forever to get a free Mac runner. It's hard to imagine how another CI service can be any worse TBH.

I thought CircleCI offered iOS builds which neither Github or Bitbucket did in the past (maybe they do now?)

GitHub Actions definitely offers iOS builds.

GHA has macOS runners but they're persistently 9 - 12 months behind macOS versions, making it difficult to keep using them when Apple starts requiring the latest OS version for Xcode. Luckily self-hosting is pretty easy, if you have the time, money, and expertise.

Agreed on their stability. For a period of time earlier this year it felt like CircleCI went down every other day, and since we require checks before deployment it was a pain constantly having to deal with outages.

Atlassian's bamboo has a terrible reputation

You're being downvoted for telling people exactly why they got laid off. It's time to wake the fuck up engineers. When you complacently don't pay attention to your business and core offering, this is what happens.

Engineers aren't generally setting the priorities on what gets fixed when or what the business is actually offering. The complacency problem is usually in management.

Can't speak for how CCI is run, but that's how it is at most places with more than 10 people.

Exactly this. I've worked for founders at small companies (<20 people) who just did not want to listen to advice or suggestions. It would be almost impossible to change anything at a larger firm. Let's just keep doing the same old thing that isn't working.

I think the saying is “vote with your feet”. There’s always at least one thing you can change. Your time is worth more than your salary, otherwise they wouldn’t pay it. If you don’t think a business will survive, stop investing into them with your time.

Definitely. And I did exactly that, eventually. (I waited too long!)

Engineers choose who to work for.

We go through these cycles where “I have no idea how they make money but they keep paying me” stops working. There’s only so long you can work for a company that doesn’t have a viable business strategy.

I don’t expect engineers to fix the business strategy, but I expect them to consider it when choosing to join a company or to stay.

> I don’t expect engineers to fix the business strategy, but I expect them to consider it when choosing to join a company or to stay.

For many companies, you cannot be qualified to make this determination.

Let's say I interview at a farming tech startup. They tell me that there are X million farms in america, and Y million have told them they want the crop software they're building. How do I make that determination of whether this is a viable business strategy? I'm not a farmer, I do not know enough farmers operating large farms to gather that data myself, I have to trust the company to represent this truthfully.

This even applies to things like CircleCI, where the product is something an engineer can understand well. I know what tools I as an individual developer use, but CircleCI is targeting enterprises, which I decidedly am not. I have no clue how some enterprise shop works. Again, I have to trust how the business present itself.

I do agree, though I think on the CircleCI front at least, it's had some not positive press on Hacker News this year (big price rises and or cutting the free tier from memory, plus the launch of GitHub actions).

But yeah, for industries we don't understand... hope it's publicly listed and has some insightful annual reports is probably the only option.

Engineers should be domain experts. This isn’t the case now but I imagine as more people become engineers, and more stuff leans towards automation, this will end up being the case

> For many companies, you cannot be qualified to make this determination.

It’s hard but useful. If you know this, you will be more successful whether you’re in engineering or sales or whatever.

I research every business and organization I do business with. It’s not perfect, but it’s part of my decision making process.

In CircleCI’s case, this would be me looking at the financials (hard because they are private), talking to some friends who use them, and, since I know some tech, trying the product.

It’s not wise to trust every business as every business has people thinking and saying they are great. It’s wise for an individual to assess these for themself. Companies success isn’t entirely random.

This research helps you form a more informed opinion, but it doesn't make you any more qualified at predicting the companies future success potential. The people running these companies are barely in such a position, nonetheless the average person who thinks they're privy to prescient insight.

I’m not sure what you mean by qualified.

The research makes you more able to make an accurate decision. It helps improve your outcome.

I don’t think being qualified is important here. I think the important piece is to make a good decision to join a firm, or leave a firm.

There’s no perfect information, but every potential employee has the ability to improve their chances by researching important decisions like this.

I think a good litmus test is if I can’t answer “how do they make money” or “how will they make money” or “how do they create value,” then I don’t want to work for them.

Instead of just giving up that you are not qualified to do so, you can at least do a preliminary analysis.

1) Follow the money - where is it coming from ? why would it keep coming in, what is needed ? where does it go ?

2) Competitors - who are they and what are they doing.

For your hypothetical farming tech company, find out about other farm tech companies who have similar product/market segment and look at their offerings, valuations and revenue model. If one does not have this reference point and doesn't have the domain knowledge, then one should not consider said opportunity unless one wants to gamble.

> Engineers aren't generally setting the priorities on what gets fixed

A good engineer would be like: “f—k it, I fixed it.” And the priorities would then get shifted around it.

Identifying when this is the right time to act and fixing it often makes the engineer gain seniority.

(Yes, it only works when both business and tech are broken. If it’s just the tech broken, then engineer will probably get pipped for working on wrong thing)

I'm not sure how an engineer could be like "f--k it, I fixed it" to the company's business and pricing model.

By being 10x, full-stack, and a ninja rockstar.

I've done this multiple times in my career.

The context to my post are the reliability issues that def. can be instrumented and fixed.

Business and pricing model not so much unless the engineer is willing to add the Sales prefix to their engineer role ;)

But even the pricing model - measure and compare the costs, competitors, and value derived (ok, commodity now vs 5yrs ago) and work with people pitching prices.

That’s still designing, analyzing, measuring systems - prices instead of code - definition of engineering.

I think the point is more that there's increasingly no reason for CircleCI to exist. All your major code hosting players have a CI tool and the various accessories you'd want to attach to it.

I think CircleCI has a niche in orgs that self-host code in e.g. phabricator or gerrit, and want a cadillac CI experience without building it themselves in jenkins or similar. I'd argue that niche is shrinking, and as a CCI employee or potential job-seeker I would wonder whether CircleCI is a good place to be because of that.

Even that niche has better competitors, I think. I haven't used on-prem CCI, but I used hosted CCI for several years and when my team switched to https://buildkite.com/ it was a huge breath of fresh air. I think BuildKite is the only CI system I've used I thought was actually worth paying for, and I bet it works out cheaper than self-hosted CCI in most cases as well.

I never used CircleCI so I can't speak to it specifically, but there's tons to improve on GitHub actions.

The Actions web UI is so bad it's almost a parody. Debugging stuff is harder than it needs to be. No good way to manually control builds. Possible but not very easy to run stuff locally. Very limited platform support (Linux Windows, and macOS, although other platforms are possible with VirtualBox hackery but it's SLOW and pretty unreliable).

It is enough to make a viable business when GitHub actions is right there? Who knows... But there are tons of reasons for external CI tools to exist, IMO.

We intentionally involve our engineers in product decisions and customer discovery.

We don't expect them to be driving novel, new features (unless they're technical in nature), but we do expect that they call out BS and make sure what they're working on is valuable to the business. Exception is platform teams who's "customer" is other developers.

> exactly why they got laid off

not really. A circleci pm down a few threads mentions its not caused by any available numbers. Another observation is that currently even perfectly healthy, focused and money-making businesses with solid products (e.g. stripe) also do layoffs.

As for circle, it's ok. I switched our thing to it from internal Jenkins, it has its ups and downs, not the worst in my experience.

No, I think this is short sighted management not realising how hard this will be for them to recover. Hiring is a hard, slow process - they won't recover this scale quickly.

Not being able to find work for engineers is a failure of imagination - again a failure of management.