Hacker News Reader

Comment by mwj | Hacker News Reader

Give me service availability over "source availability" any day. The gitlab paas product has shocking performance and uptime.

View on HN

Why not have both and host it yourself?

Maintaining a GitLab server takes time.

This is just utterly wrong.

yum upgrade

Done.

No problems for years now. Not a single one.

Are you using it? I've yet to come across anyone hosting their own gitlab instance that has not had trouble of some sort.

I am using it on Digital Ocean. I have had no issues whatsoever. Not only that, it’s a 1-click application too so I didn’t even have any setup issues.

That’s not to say I won’t have trouble of some sort one day - but what software is 100% trouble free?

Edit: typos

Really? I've been using and selfhosting it for years without issue.

My company has been self hosting the CE edition for quite a while; it's been totally smooth. Zero issues.

Same here. After a few years and moving it around from server to server (switched hosts a couple times), it's run flawlessly and upgrades are just one apt-get upgrade away...never had a single issue with that. We have 40+ projects and utilize container registry and CI with Kubernetes as well.

Dang. Afaict, you guys are the exceptions to the rule.

I guess you folks gave it enough hardware to run properly. From what I'm reading that's what's missing in a lot of failing installations. It seems particularly resource hungry.

Code repository management is a central part of every company. I don't see any problem in throwing sufficient hardware at such a crucial piece of software.

Relevant: https://docs.gitlab.com/ee/install/requirements.html#hardwar...

Thank you. I wonder why the advise a minimum of 4GB of RAM and offer a build for Raspberry Pi.

Changing that URL from "/ee/" to "/ce/" gives the same RAM requirements.

I would be guessing someone at gitlab was playing around with his/her Raspberry. But yeah, I would advise against running anything important on a Raspberry ;).

It's currently running as an omnibus installation on a KVM node. 8GB of memory and 4 vCPU (E3-1230v6). Anything more than 4GB of memory and a couple cores appears to be sufficient for our use.

Actually, one of the things we're liking about Gitlab is how lightweight it is.

...of course, it replaced Upsource for us, and Upsource is incredibly resource hungry, so maybe we just have a very, very skewed frame of reference. :)

Anecdotal but I’ve gone through two major versions and a bunch of point releases at $dayjob and the worst that ever happened was a failed database migration that didn’t actually break anything.

It’s probably the easiest thing to upgrade in our entire app stack.

I ran the latest version inside docker for 6 months and used daily CI/CD across multiple environments. It's been perfectly stable and fast. Highly recommended.

I've used it at two different jobs where we hosted it ourselves, and I host it for my own projects at home, and I have had zero issues with it.

I'm actually surprised that many people you know have had issues with it.

Self hosting typically lowers service availability, rather than increases it. For example, how many availability zones are you putting it in?

Reply on news.ycombinator.com