GitLab 15.5 | Hacker News Reader

GitLab 15.5

about.gitlab.com

83 pointsjbk86 comments

I was a happy gitlab user for a long time. As ops manager I led the migration from GitHub to gitlab 3 years ago. At the time it seemed like a great decision. Now, as tech director I am looking to switch us back. It is very clear in my interactions with gitlab they do not care about small companies. Their pricing structure is bonkers and has completely catered out the middle. IMHO the jumping the shark moment was the IPO.

I want gitlab to succeed. I’ve spent several sessions with various teams there sharing my feedback and it goes nowhere. I am not arrogant enough that I expect it to be implemented; there are no follow ups even for clarity or to say “thanks but no thanks”.

How is their code search so broken at this point? What features are they cranking out that are more important than making the site usable? Right now it looks like they are solely focused on implementing pricing increases, storage limits and user restrictions. I have had a representative say to me why don’t you upgrade to premium- as if a $99/seat/user/month is palatable to any company or a realistic solution.

I get that companies need to make money, but their approach is tone deaf. They had an issue where pricing was discussed and it ended up changing nothing [0].

I agree with the sentiment in this thread. It seems there are two common themes. 1- the inconsistencies in the UI are frustrating many people. 2- The pricing that probably made them an attractive alternative is no longer there.

I look forward to the incoming PR damage control in this thread per company policy [1], that again will lead to no changes.

[0] https://gitlab.com/gitlab-org/gitlab/-/issues/213185

[1] https://about.gitlab.com/handbook/marketing/community-relati...

The elimination of the basic tier is the thing that really threw me off.

I want GitLab to succeed because I want there to be a viable competitor to GitHub. But they've leaned so hard into this "everything in one" shtick that it's now impossible to just pay for them as a GitHub competitor. Their new price point only makes sense if you're also using their issue tracker and doing crazy things with CI that can't be done with GitHub actions.

Maybe there's a market that's interested in this all-in-one stuff, but I suspect most companies already have an issue tracker and are not going to invest in switching, because that causes headaches for years. As for CI, there's definitely a "hardcore CI" market for whom GitHub actions is insufficient, but they already have a solid open source option in Jenkins. I worry that GitLab's lost focus on the git-with-simple-automations market will cement GitHub's dominance for the foreseeable future.

We use it as an all in one but even at that price it doesn’t make sense. The problem with being all in one is that you at least have to execute at everything. I really feel they’ve lost their way in implementing the meaningful things. It’s CI is the only thing that’s keeps us going.

Same here, very unhappy customer. The new pricing is just nonsense, and now I'm looking for alternatives but migrating out 500 users is not trivial.

Moreover, I'm not too fond of the GitLab approach of replacing all the other typical tools in the DevOps toolchain. I'm happy with Jenkins, artifactory, etc., and I don't understand why I need to pay for something I'm not using or planning to use.

Is anyone in the same situation? :)

Yeah, if you want anything you need to build it yourself and raise merge requests and hope they merge it to make it into Gitlab.com.

Even their Gitlab.com runners are still not on par with the Github. You pay $99/month/user and can't even get the same base runner spec as Github instead of you need to pay double the CI minutes to get it. Pretty mad

To be fair here it’s a bit apples to oranges.

Gitlab is optimised more towards running your own runners, it’s natively supported and extremely easy to set them up.

GitHub on the other hand provides you CI which has the possibility to self-host, but they’re sort of awkward to use due to the fact they force you to use odd tags and you don’t get all the features.

Doesn't mean they could have the same level of runner spec as Github.

Especially, when you pay a lot of money for Gitlab. Github std spec is ~7gb + 2vcpu while on Gitlab you need to pay double minutes to get that. They could at least match that for their paid plans.

Personally, I think, Gitlab keeps adding a lot of features but the core CI related functionality seem to get less attention these days.

Wow, lots of gitlab hate here but at thee same time lots of love. Everyone loves gitlab. I think the core issue is when things become so complicated, you can run your own git server. As I mentioned in another comment, projects like gitea are designed for complicated problems like this. If you are as savvy as you seem to be, spin up a VM and run gitea. A normal VM at digital ocean at $1k a month could handle any conceivable load. Don't 3rd party something you can do yourself for free.

Has anyone ran gitea at a company?

I’m pro open source and run gitea at home, but looking at their issue list and pull requests it doesn’t really come off as a mature open source project that I’d want to run my company from, when down time at the wrong time can get really expensive.

In my experience: The UX, performance (objectively: not just in terms of CI speed, which is poor, but Largest Contentful Paint times for most user-facing pages on gitlab.com as well as the average API response time figures for e.g. listing issues are also poor), stability and "community" (I'm thinking of third-party apps, integrations and marketplace for reusable CI components) is strictly worse than GitHub.

Basic accessibility features for colour blind users lagged behind GitHub for years, making it difficult for me to even review diffs and work out what had been removed/added. In the end, I think the customisation feature only got shipped because an external company opened a merge request - and parts of it still fail WCAG AA today.

There's no longer a gap when it comes to GitHub and GitLab in terms of CI support, and I really don't think some of the e.g. Kubernetes integration functionality justifies picking it as the canonical location for an organisation's code.

Recently they've had to start to cut back on storage limits (and had to 180 on deleting repositories to try and free up space), which casts doubt in my mind on their financial stability.

Really though, I'm just fed up of having to put up with using it day-to-day, because clients I work with decided it would be great to pick it X years ago when GitHub wasn't as feature-rich. I now look at companies who cheap out and run self-hosted GitLab/GitLab.com with some suspicion, and question why they're not using the de facto standard tool for the job. I look at it the same as if I found a Java shop that didn't use JetBrains' products for development, it's a red flag.

> I'm just fed up of having to put up with using it day-to-day, because clients I work with decided it would be great to pick it X years ago

This is my sentiment today when using the product.

My own anecdote: I took my most recent gig because of them self-hosting their Git forge and preferring open-source and indie tools to the ones put out by megacorporations.

I'd see lock-in to GitHub as red flag instead of an “industry standard” requiring its employees to give away data to Microsoft, and making all of their code feed the Copilot machine.

That's a reasonable position to hold, but if the goal is FLOSS I'm not sure quasi-proprietary GitLab (the minute you go beyond the basic features in the free version) is much better than GitHub. Some folks have concerns around Copilot but in most corporate settings, employees are going to be using private repositories which are specifically excluded from the training corpus.

I think I'd have more respect for a company using SourceHut or Gitea if their priority was avoiding non-free services. I wouldn't personally filter a company on this basis (though I do have strong views on the operating system I'd have to use), but the demand for software engineers means it's quite possible and I respect folks for being principled.

I prefer it because even with the UI issues, its still better than GitHub.

IMO, they are trying to sell a complete end to end experience for your entire company with the aim of replacing a lot of software in your stack with a single integrated system.

IMO they need to do a better job of communicating that for the price to make sense.

$100/month is nothing for tooling critical to your (hopefully) well paid staff's productivity. Why don't you pay more?

If the choices were "use no tooling" or "use GitLab at $100/mo" I would agree with you, but GitHub offers essentially the same enterprise-level tooling for $21/mo and has a $4/mo tier that covers most people's use cases. If anything, as others have noted, GitHub's tools are more polished and more reliable than GitLab's.

Why GitLab thinks their product is worth 5x as much as GitHub's is the question at hand, not whether the price point could be acceptable in a vacuum.

I mean the big reason to use gitlab is to self host and/or use their CI/CD tooling. Doing either with GitHub is obscenely expensive and error prone. So I can see why Gitlab is worth more.

Either way, $100/month is nothing. Talk to EEs about how much their tools cost. Software developers are incredibly privileged about paying for anything. If you're spending 10k/month per developer anyway at least, what is another $100?

It's not "nothing". When you have to pay that per team member, it adds up to a vast sum very quickly. Especially when you're required to pay that for everyone using GitLab, even just to use issues or download build artefacts. That's a complete rip-off.

If you consider whether that is delivering value for money, well that's for each team/company to consider. For some I'm sure it is. But for many, and I include myself in this category, it's not value for money given the features and services delivered for the eye-watering cost hikes. It's also not competitive with what other companies are offering, and most of it can be achieved with discrete tools which are of much better quality and which cost less. Their goal of being a total integrated solution is fine--but I'm not sure most people actually wanted that--and that large breadth has come at the expense of depth and much of their offering is better served by more specialised tools. They need to refocus on the basics people want and will pay for, and be competitive by charging the going rate for that instead of 5x the true value.

I'm not averse to paying for what I use. I personally pay for all of the JetBrains products along with some other subscriptions. And my employer pays for other stuff I use for work use. I used to pay for GitLab, but I dropped my subscription because they were asking for too much. The cost has to be commensurate with the value delivered. I get a great deal of value from my JetBrains subscription which I use constantly. GitLab is not currently remotely competitive. Fix that, and I'll resubscribe.

Because its $100 per month PER USER.

Its not worth it. Spare me the sanctimonious "you pay a lot now why not pay more?" - because I'd rather not pay more for something that is not worth it!

Didn't know [0] had been locked. The bullshit edit about creating new innovations to help customers derive more value is pretty much what I expected to come from this.

You really hit the two big problems with GitLab which I wish they would address. But guess they are making enough money they don't care. Sad.

I'm hoping for a gitlab release with no new features just UI optimizations. It's a great product with a huge range of features but it feels like a kitchen sink of tools glued together. The UI is inconsistent, some parts dynamically update some don't. The settings interface is a mess. Merge requests review UI is nowhere as good as github. I wish some time was spent on polishing and ironing out these details.

Every time I have a problem or miss a feature, or it behaves in an unexpected way, I check if there's an issue open. Most of the time there is, multiple years old, sometimes with a bunch of missed release goals.

With all the recent pricing changes and the way billing if set up, I can't help but feel their end goal is ticking boxes on as many "enterprise" shopping lists as possible by adding as many features as possible to provide a nice resume, but without actually maintaining those features.

Smaller teams are becoming increasing irrelevant, and when you call them out on it and they say they do care, you just get pointed at a sales representative whom you never hear from again.

That matches my view pretty well. They support "everything", but most of that support is half-baked, just enough to satisfy a box-ticking exercise.

Their runners are nice, but the whole devops stuff and other things seem kinda half-baked and barely coherent.

Merge requests review UI is nowhere as good as github

This is my main issue with it. I don't really care about the settings UI: mostly using the API for that because I have a couple of projects which all need the same settings, and otherwise it's not visited often. Code review on the other hand is super common, and its UI in combination with that UI also not being very fast is quite annoying.

What specifically about the code review UI is annoying you?

I honestly find MR in web interface has sucked in every platform I've tried.

Use the `lab` commandline tool, it's good and makes managing MRs really easy.

makes managing MRs really easy.

How does that work? Watched the video, checked the docs, I don't see any MR features?

Upon searching I find there are at least 4 popular commandline interfaces, 3 of which are called "lab". Sorry.

I'm pretty sure we use this one: https://zaquestion.github.io/lab/

Ah, that explains a lot.

I'm pretty happy with the general GL experience but some of the features are just useless. I tried out the Sentry integration and found it to just be a worse version of the index page on Sentry, not really any point connecting it to GL. And then other things like deleting old images in the docker registery seems to be just broken to the point I couldn't work out how to delete them. And then the usage quotas were being brought in while I still couldn't clean out these hundreds of old containers.

> Merge requests review UI is nowhere as good as github. I wish some time was spent on polishing and ironing out these details.

Other than the partially dynamic updates, what do you dislike about merge requests compared to GitHub pull requests? Is it a general layout/UI thing or is there some other functional difference?

I don't have it in front of my right now to go into detail but one thing I find very frustrating is functional. When there is a merge conflict its a 50/50 coin toss if the conflict can be displayed in the UI or not.

Do you mean displaying conflicts directly in the diff? At the moment I believe an alert is shown on the file if there's a conflict, and the comparison changes to the source branch and the base of the target branch.

Depending on the complexity of the conflict, you are able to resolve via the conflict resolution UI (otherwise you'll have to just resolve it locally).

Correct, conflicts are not shown in the diff just the fact that they exist.

Wow, rewind 2 years and the comments for any GitLab post looked really different. I'm still a GitHub user/fan - but seems the GL folks might want to pay attention.

It’s very interesting to see how different the responses are. I think if I was in charge at GitLab I would at minimum be trying to determine why.

In the last 2 years GitHub has improved massively while GitLab seems to have only hiked prices while not improving much on the reliability and general UX sides. I don't personally have any issues with them putting quotas on all the stuff that used to be unlimited because I recognize things have to remain sustainable, but to justify these costs, the service has to feel rock solid like GitHub does.

Our internal gitlab spends an hour every night intermittently responding and at least 20 minutes literally unresponsive because of running the “backup” and “upgrade” steps.

I’m told that we’re talking to the Gitlab premium people and it’s currently unavoidable. We don’t have a large instance. “Maintenance mode” makes the whole process slower, so we don’t turn it on.

And I think premium now costs the same as self-hosted GitHub?

Has anyone had similar bad experiences with medium company (~500 employees, probably 1/4-1/3 developers) enterprise self-hosting?

Have you tried gitea? It's super hard core. Also, it's written in golang.

The only thing that will cause gitea to fail, as far as I can tell, is the hardware. It's basically a daemon on top of the linux kernel.

Stand on the shoulders of giants.

When you can.

even the gitea people are using github for development. The reason they mention on the migration issue is, that not all features needed are ready. They e.g. implemented issue deletation a few months ago.

https://github.com/go-gitea/gitea/issues/1029

>even the gitea people are using github for development. The reason they mention on the migration issue is, that not all features needed are ready. They e.g. implemented issue deletation a few months ago.

Actually,data gravity is the issue, not features. https://github.com/go-gitea/gitea/pull/18165 needs to be merged to be able to leave Github.

They are writing an import function im this pull request.

I think the code an pull requests are easy migrated manually. The 2k open issues would have to be cleaned up first. Then migrate the remaining issues via api. One issue could be the non existing user on the import side.

"data gravity" is itself a useful term but I maintain irritation at the person I believe coined it because they presided over the death of the basho era.

Is there another pithy term that means this that I can use?

Gitea has had reliability issues whenever I've tried to use it. And by "reliability issues" I mean the thing would spit out an "internal server error" message if I clicked around too fast in the UI while it's running locally on the same PC.

It is a lot faster and lighter than Gitlab, but I don't really care about that for something like this. Gitlab is fast enough, and since I run it on a dedicated server, I don't really care how much memory and resources it's using as long as it doesn't exceed the limits of the server.

EDIT: to be fair though, I haven't used gitea in at least a year. Things might have improved in that time.

Presumably this is only the SQL backup? (you can backup almost everything else using a regular backup system)

20 minutes seems really long for running an SQL backup. How big is the database? Is the disk/storage fast?

Have you tried replicating the database and running the backup off the secondary server?

Restarts are annoying though, and fairly slow (2-3 minutes?), but less than once per week?

Maybe using filesystem snapshots (either directly on supported filesystems or using LVM or the like) could remove the bottleneck here.

Are you monitoring the system status? How’s the CPU and I/O load when the backup is running?

My email is in my profile if you have an open ticket number for this issue that I can take a look at when I'm at my desk Monday.

My company is forcing my team (data science) to shut down our gitlab server and move to azure devops (ran by software engineering). They’ve configured ADO as a pile of shit. I want to quit.

Wasn't azure devops deprecated yet after github aquisition?

ps. I'd probably quit as well

haven't seen an official announcement of this, only various people assuming in various places on the internet.

if you have any official statement to support that i'd be happy to see that.

Nope, it and TFVC (an alternative versioning system to Git) are unfortunately still alive. But at least Azure DevOps supports Git, which makes it slightly less horrible.

And it doesn't look like it's going anywhere, with Azure DevOps Server 2022 RC1 being released in August (https://learn.microsoft.com/en-us/azure/devops/server/releas...).

Did I mention how much Microsoft sucks at naming things?

TFVC is more like Subversion but has a lot more features. It has next to nothing in common with git except that you can use it to version control files. Although I would agree that git's workflow is better for almost all use cases.

Azure DevOps has a lot more features than GitHub for micromanagement which means that middle management loves it. You can design elaborate workflows on boards and link almost everything together. It also supports reporting, connects directly to the Azure AD, and supports complex permission management with ACL for nearly every feature you can imagine.

It also supports free stakeholder accounts.

One of my clients use Azure DevOps and I hear nothibng but praise. MS Teams tho, not so much.

Also I think AzureDevops just got a revamped "Sprint page 2.0" from what I heard so they are releasing new features.

Azure Devops is fine.

FWIW, the Software Engineering team might not actually like Azure Dev Ops. Misery loves company.

Teams can get stuck using platforms they hate (e.g., MS Azure Dev Ops, MS Teams, MS Exchange and Outlook, etc.), but not wanting to take the interruption and effort hit to switch to something else.

Or it might not be their choice. Organizations can also get stuck because, say, some non-techies made the purchasing decisions, staffed support people with familiarity/certifications with those, committed to contracts, etc. Those people are politically invested in that being considered a success.

Even someone who isn't politically invested would be in the position of calling out bad choices, which is awkward and worse. They'll also be putting their neck out, if try to propose changes (which would be corrective costs no one wants, and also might fail for reasons other than their merits).

(I'm talking bigger-company politics, because that's who tends to buy into MS enterprise sales. But if you're actually a tech software startup that's inexplicably using things like Azure Dev Ops, find out why, and get that person the compassionate healing help they need, while shifting the startup onto a more sensible and respectable path.)

> Email notification when two-factor OTP attempt is wrong

I love that idea and hope I never see one of those emails

Separately, that's awesome about them upstreaming fixes to kubernetes, but latest comment on the issue that the PR has tagged as closing totally captures my experience with all those stalebots:

> @k8s-ci-robot added lifecycle/rotten (Denotes an issue or PR that has aged beyond stale and will be auto-closed.) labels 8 days ago

> Email notification when two-factor OTP attempt is wrong

Sounds like they read HN:

https://news.ycombinator.com/item?id=32911299

Anyone knows when the limit of 5 free users on Gitlab.com (SaaS) will be implemented / effective ? https://about.gitlab.com/blog/2022/03/24/efficient-free-tier...

It was supposed to be out on the 19th but it's the 22th and I still don't see any difference

GitLab team member here.

We intend to roll out the application of these limits gradually and impacted users will be notified in-app at least 60 days prior to the user limits being applied.

If you have not yet received a notification, you have at least 60 days until you will see any impact related to this change.

Since you're here commenting: the fact that you guys eliminated the $5 per user per month tier before making this change meant my company has started switching to GitHub. The combination of the two changes makes it feel like you really weren't interested in converting free tier users to paying customers so much as in shedding freeloaders, which is unfortunate.

We on the dev team would rather stick with GitLab, but "Microsoft is icky so we need to pay four times as much for the same product" isn't very persuasive to the people holding the purse.

Maybe we are simply not meant to be their customers ? No need to suppose they wanted our money in the first place

Thanks John ! Where will the notification be sent exactly ? By emails ? As a todo https://gitlab.com/dashboard/todos ?

I believe it will be done via a banner in the app but I’m looking to confirm this.

I will follow up if/when I can provide more specificity but it might be Monday before I have any more information.

Edit: the notifications will appear as global alerts directly under the Nav for all group owners.

It amazes me GitLab hasn't improved in any way their issues, projects and milestones UX.

I'd like to move off of GitHub but I'm disappointed that I'll have to leave behind the discoverability and cache of GitHub. I'm not sure that it will be as clear if I put a GitLab on my resume, for instance. I imagine others have given this thought; any advice or experiences to share?

I started on GitHub, moved to a job that used Gitlab, and then another that uses Bitbucket. Being platform agnostic about CI/CD and Git, I think, looks better than being wed to a single provider.

Think of this the other way around. Do you want to work for a company who equates "GitHub" with coding and doesn't know what GitLab is? I wouldn't.

I think lots of perfectly good companies have hiring people who might make this kind of mistake. But I'm also concerned it would doom an open source project, as contributors/users would have more trouble finding it.

I agree. There's a popular roguelike which is hosted on GitLab (called Infra Arcana) and it's not always easy for people to find it, nor to find the equivalent of GitHub's Releases page.

What discoverability does GitHub provide anyway? I know they have that Explore page but I was under the impression they were trying to kill that. I have mostly found projects through the source link on whatever package manager I discovered it on.

What are people using these days for self-hosted version control in a corporate setting, Gitlab/GitHub/gitea? I’m currently testing out Gitlab for deployment at work but finding it frequently goes oddly unresponsive. Is gitea appropriate in a corporate setting or is it more for hobbyists?

Depends on your needs. Do you only need a remote repo and issues + pull requests? Gitea all the way. Do you also do project management (sprints, milestones etc) and want CI/CD and integration with things like k8s etc? Gitlab (Premium)

Gitea for code, docs as code with Woodpecker for CI for test tasks and build/deploy tasks. Also integrated with Weblate for translation.

Kanboard for more complex task management with a hook server for custom automations.

GitLab internal instance and GitLab hosted.

I like it, though we probably could do permission automation better on the internal instance. The team who run the hosted stuff have this perfect.

Phabricator at my company. It's surprisingly good. The main downsides are no built in CI, it's written in PHP, and it's been abandoned. There is a community project to take over but too early to see if it will succeed.

I would probably use self hosted Gitlab or Gitea/Gogs if I was choosing now.

Gitlab is the Firefox of browsers.

You mean it's not backed by evil corp? Yeah.. I think you're right.

Run by people who constantly make decisions that are contrary to what its community and supporters want?

Gitlab is totally buggy and way too expensive.

Hope it has better noscript/basic (x)html support, at least for core functions (like bugzilla has).

Did you file an issue about that experience? I would not expect any such thing to be prioritized but it for sure won't be prioritized if there's no issue for it

Yes, a long time ago I did manage to contact somebody there, got the finger.

I don't expect less from most javascript kiddies.

It's not an "issue" its the design. GitLab is a web app and not a document so there is no chance they will rewrite the whole thing to support not running scripts.

I think you're lobbying the wrong user; I'm not the one trying to noscript the Internet :-)

I was just reminding the commenter that bitching on HN is not actionable. Neither is "contacting somebody there," but at that point, I had done my part

Reply on news.ycombinator.com