The "Cloud" is great until it's not
This is why on premise installations will never go away (hopefully) as an option.
Bitbucket, Atlassian's hosted software version control service, has been afflicted by a series of service disruptions over the past three days, to the predictable consternation of programmers. Git and Mercurial, separate software version control applications supported by Bitbucket, began misbehaving on around 2241 UTC on …
On premises can have the same problems, server goes down mysteriously and needs to be troubleshooted for days. The main difference is that your organization has to/gets to do the fix themselves which can be good or bad. I'm not saying that "cloud" solutions are better, you're totally reliant on the cloud provider, but on-premise solutions have similar reliability problems.
Of course it can, but when it's the cloud provider, it's all of that provider's customers that have the problem. When it's your on-premises server that has a problem, it's only your customers that are impacted.
Let's try avoiding humongous single points of failure, shall we ?
On the other hand, on premises loads on a service are rarely quite as much as loads on a cloud-hosted service, so fewer load-related problems are likely to occur.
On the other hand (I'm running out of hands), a cloud service provider may well have more in-house expertise. Or less!
There's pros and cons for both. Of course if you're stuck with a PHB who just follows fashion, you're fucked in cloud or on premises.
> On premises can have the same problems, server goes down mysteriously and needs to be troubleshooted for days .. but on-premise solutions have similar reliability problems.
I beg to differ, no they don't troubleshooted for days, swap-out the hardware and roll back to the last backup. Competent tech people usually know the local set-up better than their own back garden.
"At Atlassian, ensuring system performance and reliability are of the utmost importance to us," a spokesperson said. "We are aware of the issue and are providing real-time updates to customers via the Bitbucket Statuspage."
Spokesperson? SpokespAI?
These chat bot systems always seem to say the same damned thing :(
I mostly manage to avoid our corporate Confluence install, in favour of putting docs into our team's GitLab instance. Sadly I can't completely avoid Jira, Confluence's smelly cousin.
Atlassian should be some kind of case study for what happens when a small company who "gets it", gets big and starts selling "enterprise" rubbish.
To be fair; Confluence & Jira are pretty good as far as commercial Wiki & Ticketing systems are concerned. I know that's not saying very much, The the alternatives are so much worse.
I can't really understand how people can make software so truly awful; To the point you can actually *SEE* the effort that the put in to make it bad :(
I was wondering about that. Do you suppose there a degree in Spokesperson Speak, where you learn how to say everything by saying nothing? Either that or there must be some special iOS SpokesPerson app that takes "We are really, really sorry but we have not got an idea of what is happening or how to fix it." and translates it into, "We are aware of the issue and are providing real-time updates to customers".
Politicians take the next level of this course. They appear to say things, that are what you want to believe. When those things don't happen, on later examination of their words, they didn't actually say those exact things - just close enough to allow your biases to believe they did before the election.
Bureaucrats are somewhere in the middle level of competency at this game.
"Oh, piebald, you had nothing to say, yet you made it into words and said them anyway!"
-- CS Lewis, Perelandra
That's still true for bureaucrats, but from what I see, although they used to, politicians these days don't bother to employ that level of subtlety. They'll just say exactly what you want to believe (including mutually-exclusive things to different people, if necessary). When those things don't happen, you can later examine their words, find that they really did say those things - and then cry a little when you realize that this doesn't really affect elections much, if at all.
What's great about git is that it doesn't really matter if the remote repo becomes unavailable for a while because you've still got all those commits in the local clones. Sure it may take a little while to get yourself out of the bind but it is ultimately doable even if the remote repo can't be recovered at all.
It may be a bit inconvenient to be unable to share updates in a team in the normal way but there are simple ways around this (e.g. using generated patches and the like).
I'm not saying I want downtime from that cloudy git, but actually there's no real chance of missing those pulls or losing commits unless you are a complete noob and delete or mess up your clone.
Indeed the beauty of distributed vcs is that you don't need a server, ever. Just nominate a local repo as the authoritative one and sync with it. That is the way it was meant to be done in the first place, Bitbucket, GitHub, etc are trying to force dvcs into the classic vcs mould.
If you've made your production systems dependent on someone else's cloud then you should have prepared for that cloud to be not available. I find it amusing how the people who were stupid enough to make their IT systems so dependent on services and companies whose technology and processes they know so little about are the ones who publicly whine about it on social media as soon as there's a problem, thus loudly announcing how stupid they were to do that to the entire world.
This. If your CI/CD pipeline is introducing external dependencies, then fire who created it & bring in someone competent. As I've mentioned before, I considered rubygems.org to be an unacceptable dependency risk BEFORE it was taken down for a week. Double mirrors, people.
I was speaking about my own uses. There's no point in fingerpointing when it's your own stuff, because you're the only one who cares.
Having stuff in the cloud means that I am at the mercy of the provider and have exactly zero control over what's happening, and I can't fix stuff when it goes wrong. That's too dangerous of a situation for me be comfortable with.