|
LochNessMonster posted:In this case there is only 1 correct anwser, “one of the earlier wireless protocols”: RFC 1149 - IPoAC Or where social media began and should have probably ended, finger: https://datatracker.ietf.org/doc/html/rfc742
|
#
?
May 1, 2022 14:11
|
|
|
|
| # ? Jun 4, 2024 09:59 |
|
|
E: think I figured my thing out thanks
some kinda jackal fucked around with this message at 03:45 on May 4, 2022 |
|
#
?
May 3, 2022 20:17
|
|
|
Lady Radia posted:have you ever worked with circleci? i've been considering gitlab but honestly circleci does EVERYTHING i want so loving well so far CircleCI is awful and we want to get off of it. They still have not managed to do stuff like "have a pipeline that only has a single execution at once" (without very hacky workarounds). Triggering pipelines via API is not great. Scheduled pipelines are also a bit of a nightmare. Especially re-running a scheduled pipeline has some weird edges. Their credit model is horrible and if you end up purchasing too many credits, they don't roll over to the next year, which can leave you with six figures of spoiled credits if for whatever reason your growth wasn't as you had projected. You also can't do "pay as you go" and their sales org is entirely non-apologetic about this too. The kicker is that they still do not have arm64 docker executors. Their workarounds are "well you can use machine executors, or self-hosted runner", but then you need to convert your entire pipeline to use docker-compose. And if I'm going to do that i might as well move to Gitlab or Jenkins. edit: orbs also suck, i wish they had done something more like concourse where jobs are provided as containers instead of as a bunch of bash scripts
|
|
#
?
May 5, 2022 17:31
|
|
|
It is very frustrating that there are still not any good options for hosted macOS CI. GHA's macOS builders are awful. CircleCI has much better macOS support but their core CI product is bad. Travis was good for a while but they died. Gitlab's is still in private beta. Xcode Cloud was designed to look good in a WWDC presentation rather than to be a usable product. Everything else that I'm aware of requires that you host your own builders, which is a gigantic headache.
|
|
#
?
May 5, 2022 20:39
|
|
|
Azure DevOps has hosted big sur and catalina agents I've never used them myself though
|
|
#
?
May 5, 2022 20:46
|
|
|
Github Actions uses the same build agents as Azure DevOps. Probably less incompetent sales than paying for them via Github, but they have the same problem of that there's seemingly one person maintaining macOS support as a side project and so it takes an entire year to add support for new macOS versions.
|
|
#
?
May 5, 2022 21:00
|
|
|
We have a ton of MacOS builds and we’re right on the threshold between paying the eye watering amount to have someone manage the fleet for us or to have engineers in-house maintain it. We have datacenter space and will have some bare metal needs no matter what so it doesn’t seem like a huge added expense given we could repurpose hundreds of our old MacBooks for CI and toss them out over time even. MacStadium and MacVault are both unable to meet our capacity needs it appears so we may have no choice. We can’t wait on cloud vendors to update their infrastructure normally either.
|
|
#
?
May 5, 2022 21:29
|
|
|
GitHub actions are very slow for Mac. We use a bunch of Mac minis and orca right now, it's pretty fast and usable. All the hosted solutions are so bad
|
|
#
?
May 6, 2022 01:04
|
|
|
I recently had to do some extremely hacky poo poo for a POC demo, got me thinking If you have a long lived public cluster, what's stopping someone from generating a kubeconfig for an existing service account that already has cluster-admin (or near it) and then exporting it via cell phone photo or whatever. Base64 encoded a single token based kubeconfig is right at about 1000 characters string. It's long but not that long to key in manually. These things never expire and bypass SSO. Once you're in, prod clusters often have way more access than they need you could casually pwn most companies weeks if not months after getting let go from most any VPN provider  I guess you could trail the logs for an event like this but I've never read about anyone setting this kind of alert up as a first line of defense, like fail2ban is for traditional servers Hadlock fucked around with this message at 12:38 on May 6, 2022 |
|
#
?
May 6, 2022 12:35
|
|
|
dehumanize yourself and face to Jenkins
|
|
#
?
May 6, 2022 13:28
|
|
|
Plorkyeran posted:Github Actions uses the same build agents as Azure DevOps. It's a fork. They've diverged quite a bit, but I don't know how applicable the divergence is to this particular case.
|
|
#
?
May 7, 2022 04:19
|
|
|
my homie dhall posted:dehumanize yourself and face to Jenkins After some time with Azure DevOps I can safely say Jenkins is not the biggest piece of junk in the CI/CD space.
|
|
#
?
May 7, 2022 10:20
|
|
|
New Yorp New Yorp posted:It's a fork. They've diverged quite a bit, but I don't know how applicable the divergence is to this particular case. The build systems may have diverged but the build agents are identical. When you click through the Azure DevOps documentation about what's installed on the runners, it takes you to the GitHub Actions repo.
|
|
#
?
May 7, 2022 14:45
|
|
|
FISHMANPET posted:The build systems may have diverged but the build agents are identical. When you click through the Azure DevOps documentation about what's installed on the runners, it takes you to the GitHub Actions repo. What's installed on the hosted machines used to run builds isn't the same thing as the agent software itself. It's a fork, unless something has changed since the last time I talked to the team at Microsoft. Actions has a different YAML capabilities and additional functionality not present in Pipelines, necessitating a fork of the agent software. There may still be a common core but they are definitely evolving independently
|
|
#
?
May 7, 2022 20:18
|
|
|
The agent software is not what I was referring to or the problem with GHA on macOS. It is the machines and the software they have installed on those machines.
|
|
#
?
May 8, 2022 04:24
|
|
|
luminalflux posted:CircleCI is awful and we want to get off of it. you should use the enterprise on-prem version if you can, i dont think the credits issue or invoking the api sucking is as bad
|
|
#
?
May 8, 2022 19:43
|
|
|
Lady Radia posted:you should use the enterprise on-prem version if you can, i dont think the credits issue or invoking the api sucking is as bad Docker executors are not available for self-hosted runners. If i'm going to redo my 2 kloc CI definition to fit another executor, i might as well convert it to Gitlab or Jenkins.
|
|
#
?
May 10, 2022 18:13
|
|
|
GitLab is good. do NOT use Jenkins lol
|
|
#
?
May 10, 2022 20:24
|
|
|
Lady Radia posted:GitLab is good. If it would be a bit snappier would be wonderful. On a VM with 2 CPUs (4 cores each) and 16GB of RAM (it's overkill but I just added hardware to it to see if it makes a difference) it still takes 5 seconds to display the first page. And I'm the only user of the system. Feature-wise it's fine. But drat, it barely moves. Maybe it would like SSDs, maybe it would like a real machine, dunno really.
|
|
#
?
May 10, 2022 22:01
|
|
|
Lady Radia posted:GitLab is good. do NOT use Jenkins lol Found the GitLab sales rep. It's ok, I wouldn't call it "good". The fact that it allows you to isolate your runners away from the segment that the GitLab solution runs within is nice, but the stupidity of their YAML is on-par with how bad CircleCI is if not worse. Sure you have the ability of defining workloads a bit more dynamically, but it comes with a cost. Their official documentation constantly pushes for anti-patterns in CICD workflows, and are very opinionated towards everything within the stack existing in their solution. Want/Need something that is beyond their scope, good f'n luck integrating with their poo poo successfully without opening a gaping hole in your pipeline.
|
|
#
?
May 11, 2022 00:32
|
|
|
Gitlab, for when you want to manage k8s to run your CI! ps all ci mostly suck
|
|
#
?
May 11, 2022 02:24
|
|
|
We've paid for gitlab for years, but are moving to the CE release this year because the price is going up and puts it out of our budget. How owned are we gonna be?
|
|
#
?
May 11, 2022 02:30
|
|
|
drunk mutt posted:Found the GitLab sales rep. Starting to sound like all CI sucks. I used to use Concourse when I was at CloudFoundry, which has some very nice concepts like every task being a container with 3 defined commands and you can string together fairly complex pipeline with it. However it is inherently not built for working with PRs the way literally every other CI system is, because that is not how Pivotal works
|
|
#
?
May 11, 2022 03:06
|
|
|
Volguus posted:If it would be a bit snappier would be wonderful. On a VM with 2 CPUs (4 cores each) and 16GB of RAM (it's overkill but I just added hardware to it to see if it makes a difference) it still takes 5 seconds to display the first page. And I'm the only user of the system. Feature-wise it's fine. But drat, it barely moves. Maybe it would like SSDs, maybe it would like a real machine, dunno really. Yeah, it probably needs a SSD, as I've got Gitlab running it on VM with 4 cores total with 16GB of RAM and pages load pretty much instantly. I've had zero issues with Gitlab so far and I'm looking forward to the day when I can take a shotgun to the crusty old Jenkins environment that lingers because one team keeps using it.
|
|
#
?
May 11, 2022 03:14
|
|
|
my homie dhall posted:dehumanize yourself and face to Jenkins
|
|
#
?
May 11, 2022 04:08
|
|
|
On long enough timelines, usage of any other CI platform degenerates into a worse form of Jenkins
|
|
#
?
May 11, 2022 04:11
|
|
|
my homie dhall posted:On long enough timelines, usage of any other CI platform degenerates into a worse form of Jenkins sadly true tangentially related: jira has so many features now it only takes a year or two before a junior project manager comes along, gets admin and morphs what should be a simple, effective ticketing system into a hot pile of garbage
|
|
#
?
May 11, 2022 04:12
|
|
|
It's important to stress that when anyone here is saying "Gitlab is good", what they're really saying is that it sucks the least. I certainly haven't been bit by it yet. Was recently impressed to discover that their git server helpfully told me to include a skip-ci option when I was modifying the repo from within the -ci.yml. I didn't notice until after I had already set the job to manual or scheduled only, but still. I'd rather have it let me create cursed infinitely looping jobs than anything else An SSD would help out a lot, there might also be some memory limits you can increase.
|
|
#
?
May 11, 2022 05:16
|
|
|
Gitlab also seems to really struggle on a box without a swap partition. I inherited a Gitlab server when I started my current job that had some awful performance issues that were resolved when I added a swap partition.
|
|
#
?
May 11, 2022 05:32
|
|
|
In a three node K3S HA lab, other than not being representative of (presumably) most production kubes, is there any real downside of having all three nodes deployed as servers with no workload taints? I guess overkill and unnecessary resource usage for server components might be legit reasons, but honestly if that’s it I’m happy to just run in this model. Between kube-vip and metallb I think I’d be pleased with my setup as even a two-node “server” cluster given I’m just starting to dip my toes in and my workloads are minimal (so hoooonestly, even a single node is probably fine for everything except it’s not as  )
|
|
#
?
May 12, 2022 10:25
|
|
|
xzzy posted:We've paid for gitlab for years, but are moving to the CE release this year because the price is going up and puts it out of our budget. How owned are we gonna be? You're going to miss the code search feature
|
|
#
?
May 12, 2022 11:09
|
|
|
some kinda jackal posted:, is there any real downside of having all three nodes deployed as servers with no workload taints? I guess overkill and unnecessary resource usage for server components might be legit reasons, but honestly if that’s it I’m happy to just run in this model. In my experience you don't create taints until they're needed, and you'll know when you need one As an example, originally we just had our Monolith on the cluster, and later we added data analytics to the cluster (I guess originally it was a small accessory load and didn't justify another cluster at the time) Now the analytics load is 3x the monolith, with a very different load pattern (more CPU, faster disk) so we have two sets of servers, with taints to keep the two running on their tuned machine type Should we move analytics to it's own cluster? Probably. Is it my top priority this week? Nope
|
|
#
?
May 12, 2022 19:09
|
|
|
Sounds good. I know next to nothing about these situations but I can imagine based on context that this makes sense. For my specific use I’m guessing there’s really no major harm in just having all three nodes in the cluster running server roles without being super anal about refusing to schedule workloads on anything but a dedicated worker. I’d be lying if I said I had a NEED for kubernetes, but I really like the workflow, and so far everything makes SENSE, enough that I think I might eventually either migrate my workloads to this little k3s cluster, or just reinstall ubuntu on the docker host for cleanliness sake and bring up a single node k3s. I really do like it more than just working with docker, though that’s probably more to do with the fact that I have played with kube more right now. some kinda jackal fucked around with this message at 03:02 on May 13, 2022 |
|
#
?
May 13, 2022 02:59
|
|
|
it's super frustrating that k8s lives up to the hype for the most part lol, i wish it were worse to work with and that rancher just didnt work half the time so i could argue against it.
|
|
#
?
May 13, 2022 07:29
|
|
|
E: oops
some kinda jackal fucked around with this message at 03:40 on May 28, 2022 |
|
#
?
May 17, 2022 11:52
|
|
|
Does anyone have recommendations on libs to write code that generates Azure DevOps pipeline configurations from templates?
|
|
#
?
May 18, 2022 12:50
|
|
|
LochNessMonster posted:Does anyone have recommendations on libs to write code that generates Azure DevOps pipeline configurations from templates? Maybe Pulumi? https://www.pulumi.com/docs/guides/continuous-delivery/azure-devops/
|
|
#
?
May 18, 2022 13:58
|
|
|
may or may not be the best choice but I inherited some ansible that does that and haven't needed to replace it yet
The Fool fucked around with this message at 15:28 on May 18, 2022 |
|
#
?
May 18, 2022 14:35
|
|
|
LochNessMonster posted:Does anyone have recommendations on libs to write code that generates Azure DevOps pipeline configurations from templates? What do you mean by pipeline configurations? Azure Pipelines already supports templates out of the box.
|
|
#
?
May 18, 2022 15:26
|
|
|
|
| # ? Jun 4, 2024 09:59 |
|
|
New Yorp New Yorp posted:What do you mean by pipeline configurations? Azure Pipelines already supports templates out of the box. My description was kinda bad but I’d like to generate the templates themselves by code. Notvar substitution in a predefined/static template.
|
|
#
?
May 22, 2022 19:20
|
|