Conversation
Notices
-
Disinformation Purveyor :verified_think: (thatguyoverthere@shitposter.club)'s status on Monday, 04-Dec-2023 20:34:07 JST Disinformation Purveyor :verified_think: LTS linux releases are less great than rolling releases.
"We build packages for 5 years"
So I need to re-install every 5 years?-
Disinformation Purveyor :verified_think: (thatguyoverthere@shitposter.club)'s status on Monday, 04-Dec-2023 20:38:28 JST Disinformation Purveyor :verified_think: @raucao
If "stable" releases meant table I'd be all for it. I've seen RHEL 5, 6, 7, and 8, along with numrous unbutu LTS releases break on update. It may happen less frequently (although I have my doubts), but when it does it's no less impactful. -
Râu Cao ⚡ (raucao@kosmos.social)'s status on Monday, 04-Dec-2023 20:38:30 JST Râu Cao ⚡ @thatguyoverthere Everything's a tradeoff. It's nice to never have to re-install or dist-upgrade, but it's also nice to be able to update continuously and without human interaction, with nothing breaking in the process.
Disinformation Purveyor :verified_think: likes this. -
white_male (white_male@poa.st)'s status on Monday, 04-Dec-2023 20:47:28 JST white_male @thatguyoverthere @raucao Production should have staging anyway, so that's just a scare for lazy admins. Disinformation Purveyor :verified_think: likes this. -
Disinformation Purveyor :verified_think: (thatguyoverthere@shitposter.club)'s status on Monday, 04-Dec-2023 20:47:28 JST Disinformation Purveyor :verified_think: @white_male @raucao agreed about a staging lab. I still say rolling updates are better. LTS just gives redhat an opportunity to sell extended "entitlements" to patches they still do but won't give the poors. Ubuntu has a similar model. -
Disinformation Purveyor :verified_think: (thatguyoverthere@shitposter.club)'s status on Monday, 04-Dec-2023 20:50:41 JST Disinformation Purveyor :verified_think: @raucao @white_male I have a minified version of my production environment. It is not used to test is updates although we've had updates break shit in production that could have been avoided if we'd done the work. -
Râu Cao ⚡ (raucao@kosmos.social)'s status on Monday, 04-Dec-2023 20:50:42 JST Râu Cao ⚡ @white_male @thatguyoverthere Nobody has staging for all of their production VMs, unless they run very few systems. Wanting hands-off security upgrades for all of your production environments is not lazy, but rational. Staging is mostly for testing application updates, not basic system upgrades.
-
Râu Cao ⚡ (raucao@kosmos.social)'s status on Monday, 04-Dec-2023 20:53:39 JST Râu Cao ⚡ @thatguyoverthere @white_male If i were to switch to something else in prod, it would probably be more along the lines of nixos I think.
Disinformation Purveyor :verified_think: likes this. -
Râu Cao ⚡ (raucao@kosmos.social)'s status on Monday, 04-Dec-2023 20:53:41 JST Râu Cao ⚡ @thatguyoverthere @white_male I mean, yeah, if you never want to actually upgrade, then you may have to pay. But the free tier for Ubuntu ESM is fairly generous for "the poors" anyway.
To each their own. I would never go back from a rolling distro on my local machine, but I would also never run one in production on hundreds of VMs.
-
white_male (white_male@poa.st)'s status on Monday, 04-Dec-2023 21:33:23 JST white_male @thatguyoverthere @raucao Xen takes some work upfront, but allows easy staging and virtually in place rollbacks.
Even easier with Xen based environments, Proxmox is a popular one.Disinformation Purveyor :verified_think: likes this. -
Disinformation Purveyor :verified_think: (thatguyoverthere@shitposter.club)'s status on Monday, 04-Dec-2023 21:33:23 JST Disinformation Purveyor :verified_think: @white_male @raucao I run a mixed environment (vms and iron) and as much as virtualization makes things easier it comes a trade off in performance. That said I think corporate decision makers are going to kill the physical side (a real shame considering we were once a fully virtualized environment and moved to physicals because it allowed us to run beefy database and application servers. The fully virtualized environment will probably end up costing the client thousands more per month with our pricing model and I am skeptical we'll see the performance we do today. -
Niclas Hedhman (niclas@angrytoday.com)'s status on Monday, 04-Dec-2023 21:34:07 JST Niclas Hedhman You hit the nail on the head;
Make people LTS addicted, and charge hefty fees to extend the support for enterprises that are too scared to upgrade (i.e. being on LTS in the first place).
Without fostering a culture of "being-up-to-date", large enterprises quickly get filled with old cruft.
Disinformation Purveyor :verified_think: likes this. -
Disinformation Purveyor :verified_think: (thatguyoverthere@shitposter.club)'s status on Monday, 04-Dec-2023 21:37:12 JST Disinformation Purveyor :verified_think: @Marvelous7578 @raucao @white_male we run containers on the bare metal. The problem is we're gonna have to give that up in favor of vms running containers. -
Marvelous (marvelous7578@sosial.agdersam.no)'s status on Monday, 04-Dec-2023 21:37:13 JST Marvelous LXC containers are nice when you want near-metal performance without dealing with gay-ass, nu-male Docker shit.
Proxmox's implimentation is great in my experience. You can even use VMs alongside LXC containers. -
Marvelous (marvelous7578@sosial.agdersam.no)'s status on Monday, 04-Dec-2023 21:58:17 JST Marvelous Well, that's stupid. What's the point of two layers of virtualization? (I know containers barely fit the definition of virtualization, but still.) Disinformation Purveyor :verified_think: likes this. -
white_male (white_male@poa.st)'s status on Monday, 04-Dec-2023 22:06:12 JST white_male @thatguyoverthere @raucao Depends, modern CPUs have full facilities for virtualization, 5% performance penalty is considered high on non over provisioned systems. I run compute GPUs via passthrough and can't tell the difference between my bare metal and virtualized workstations. Disinformation Purveyor :verified_think: likes this. -
Disinformation Purveyor :verified_think: (thatguyoverthere@shitposter.club)'s status on Monday, 04-Dec-2023 22:06:12 JST Disinformation Purveyor :verified_think: @white_male @raucao It depends on the workload and your level of access to the virtualization stack. I will be forced to use a shared "cloud" which means no hardware pass through or custom tuning. This after we made a very conscious decision to move to physical hardware based on experience in a similar environment.
We don't run any GPU workloads, but having our cluster on physical has been working very well for several years at this point. Being acquired by larger company turned out to make us less flexible so the option is going away. Who knew -
Disinformation Purveyor :verified_think: (thatguyoverthere@shitposter.club)'s status on Monday, 04-Dec-2023 22:39:12 JST Disinformation Purveyor :verified_think: @Marvelous7578 @raucao @white_male The point is they are shutting down the DC where we live and where we're going we don't have access to physical hardware. Really it's just going to cost the client more money because we will need to sell them many more vms to do the same work as our current cluster of physicals can do. The system is designed to scale horizontally so it won't really matter for me all that much tbh. -
Marvelous (marvelous7578@sosial.agdersam.no)'s status on Monday, 04-Dec-2023 22:41:27 JST Marvelous Reminds me of that one article about a company saving millions of dollars after moving away from cloud-stuff to running things locally on their own servers.
People, and especially bosses, are easily fooled by cloud-shit, and forget how much cheaper it is to run your own stuff.Disinformation Purveyor :verified_think: likes this.
-