|
Even in a completely greenfield environment, the "one click upgrade" promise they sell you with VXRail is utter bullshit. Don't believe the hype. Each upgrade will take a week minimum with constant handholding by support
|
# ? Nov 8, 2021 22:31 |
|
|
# ? May 9, 2024 15:33 |
|
All joking aside, Azure Stack Hub kind of rules, but I also love outsourcing all the pain to vendors and I don't pay the bills out of my own pocket so gently caress it
|
# ? Nov 8, 2021 23:17 |
|
GrandMaster posted:Even in a completely greenfield environment, the "one click upgrade" promise they sell you with VXRail is utter bullshit. Don't believe the hype. Each upgrade will take a week minimum with constant handholding by support As in there are problems requiring support to fix them with every single upgrade, or just that you've been burned often enough to want them to go with you step by step? Not that either answer is entirely encouraging.
|
# ? Nov 9, 2021 00:54 |
|
In the beginning, we'd get failures at almost every step in the process... anything from firmware bundles, to lifecycle manager patches. Just get support to deal with all the poo poo now, since we pay for support. And there's no upgrade path from vcf 3.x to 4.x either lol. Anyways glad I'm out of that job now!
|
# ? Nov 9, 2021 12:26 |
|
I just shut down my VMware homelab server because I can't justify the energy expense and it takes like 15 minutes to boot if I want to bring it up at a moment's notice to try something. Switching to an old Dell compact desktop with an i7 and a bunch of RAM, hopefully enough to run a few KVM instances. I've been out of the KVM game so two questions: is there a good minimal footprint Linux that is recommended as a KVM host? Typically I used to just throw CentOS minimal for low footprint VMs, not sure if there is anything more suitable or KVM specific these days. And, What's the hotness for KVM web management? I can create and launch VMs from the CLI, that's fine, but ideally this is a headless desktop so I need some way to interact with the VMs before they're accessible over the network etc. Just for "oh I need a windows VM for 45 minutes to try something" so I'm not looking to do a whole automated provision-to-network-available deployment pipeline here.
|
# ? Nov 9, 2021 15:37 |
|
I mean Proxmox does most of that for you and is Ubuntu/Debian based.
|
# ? Nov 9, 2021 16:36 |
|
Martytoof posted:
https://www.tecmint.com/manage-kvm-virtual-machines-using-cockpit-web-console/ I've always used virt-manager to play with kvm machines. That can do management-over-ssh, so you can run a copy of virt-manager using WSL2 or Virtualbox VM and use that to connect. Pablo Bluth fucked around with this message at 20:34 on Nov 9, 2021 |
# ? Nov 9, 2021 20:28 |
|
My prior experience with proxmox was mixed. I think I didn't give it a chance because it "wasn't vsphere" but now that I'm looking for just a barebones thing it might be a better option. All the same I think I will try to just roll my own with Pablo's advice. That's kind of what I'm envisioning, thanks. I also just literally want to think about it as little as possible, and if I didn't have the hardware on hand I'd probably honestly be better off just getting a few micro EC2 instances.
|
# ? Nov 10, 2021 15:46 |
|
Proxmox is great, I have no qualms recommending it over rolling your own KVM stuff.
|
# ? Nov 10, 2021 15:54 |
|
I am on proxmox and so far I really like it. I just have to figure out how I can use my HDDs as pure backup and put them in standby while not backing up (Which apparently means disabling the storage).
|
# ? Nov 11, 2021 11:36 |
|
I'm getting disk size warnings on my VirtualBox 18.04.6 Ubuntu VM where it says it is running out of space. I have no idea what is using all the space since I literally only use this this for social media and only occasionally, but whatever. I didn't imagine this would be an issue, since it's using a 10-gig dynamically allocated VDI. Therefore, I figured it'd grab more space when it needed it. I imagine that Ubuntu doesn't know it can do this, so it's throwing up the warning. Basically, do I need to anything? Or will VirtualBox make the drive bigger without my intervention?
|
# ? Nov 16, 2021 21:20 |
|
Magnetic North posted:I'm getting disk size warnings on my VirtualBox 18.04.6 Ubuntu VM where it says it is running out of space. I have no idea what is using all the space since I literally only use this this for social media and only occasionally, but whatever. I didn't imagine this would be an issue, since it's using a 10-gig dynamically allocated VDI. Therefore, I figured it'd grab more space when it needed it. I imagine that Ubuntu doesn't know it can do this, so it's throwing up the warning. I believe 10GB is the maximum size of the VDI disk and 10 gigs doesn't sound that much when you include browsers caches and everything else. Run the command 'df -h' to check which mount is running out of space and then 'du -hx /MOUNT | egrep "G\s"' to see what is using the space.
|
# ? Nov 17, 2021 13:27 |
|
Saukkis posted:I believe 10GB is the maximum size of the VDI disk and 10 gigs doesn't sound that much when you include browsers caches and everything else. Run the command 'df -h' to check which mount is running out of space and then 'du -hx /MOUNT | egrep "G\s"' to see what is using the space. If I understand correctly, I got this from the first one, among other things. Filesystem Size Used Avail Use% Mounted on /dev/sda1 9.8G 8.9G 450M 96% / I don't know what that means, but the numbers look right. Is that root? I'm bad with Linux. The other command gave an indecipherable scrawl of lots lots and lots of info. So, let's reverse the question: If I want to use a virtual machine for websurfing, is there a different set up I should use to prevent it from ballooning up? Either another VirtualBox configuration, or a different Linux distro or whatever else? Or could I have it have a saved state and have it roll back to an earlier state after use or something?
|
# ? Nov 17, 2021 17:09 |
|
You're using 96% of your disk. Dynamically allocated refers to the amount of space on your disk the virtual disk takes up. a 10GB virtual disk that only has 1GB of data will only take up 1GB of space on your drive. This is also known as thin-provision. You need to increase the size of that disk (inside virtualbox or whatever you're using) to say, 20GB, and then inside Ubuntu grow your partition to fill the desk. https://unix.stackexchange.com/questions/196512/how-to-extend-filesystem-partition-on-ubuntu-vm 8GB sounds about right for a pretty basic Ubuntu VM just starting out....so make it bigger.
|
# ? Nov 17, 2021 18:13 |
|
Looks like that worked. Thank you both very much.
|
# ? Nov 17, 2021 22:58 |
|
For those running Hyper-V in a large environment, are you boot from local disk or boot from SAN? We're a ~1000 VM shop and we've hit this bug where the CSV's drop every third blue moon during backups. Many fingers have been pointed, but no solutions from either side.
|
# ? Dec 1, 2021 20:00 |
|
I'm setting up my NAS with TrueNAS Scale, which uses Kubernetes. I never worked with it, and it probably has a reason why it is the way it is, but holy hell is it convoluted. If I do kubectl describe nodes, the Memory Requests column is the actual memory the container is using right now, right? It doesn't just plain echo the requests: value in the resources: section in the deployment YAML, does it? i.e. this: Also, what's k3s-server and why is it using 20% CPU while the Kubelets are doing practically nothing? Combat Pretzel fucked around with this message at 21:37 on Dec 8, 2021 |
# ? Dec 8, 2021 21:08 |
|
Combat Pretzel posted:I'm setting up my NAS with TrueNAS Scale, which uses Kubernetes. memory requests are the schedule time guarantees the scheduler uses to determine which node is capable of hosting the pod. It is not instantaneous real memory usage. limits are the upper bound where if you go beyond, the workload will be killed by the cgroup oomkiller. the delta between requests and limits represents burst capacity. burst with caution. if the node runs out of memory because multiple apps are bursting beyond their schedule time guarantees you'll get noisey neighbour issues and they'll probably just keep killing each other. k3s-server is probably the entire control plane in one binary. It's doing all of the control-loops convergence/api calls of the cluster.
|
# ? Dec 8, 2021 21:52 |
|
If this is a single host, not a cluster, do I even need something like Kubernetes? I'm "using" it because it came with TrueNAS Scale. What doesn't sit well with me is having some agent in background burning CPU for what's to me no good reason. Wouldn't Docker cover most of my bases, running a bunch of crap like influxdb and mariadb in background, resource constrained?
|
# ? Dec 8, 2021 22:27 |
|
Combat Pretzel posted:If this is a single host, not a cluster, do I even need something like Kubernetes? No. I don't know what trueNAS is, but check if they have a docker-compose version. Or literally anything else.
|
# ? Dec 8, 2021 22:39 |
|
Combat Pretzel posted:If this is a single host, not a cluster, do I even need something like Kubernetes? I'm "using" it because it came with TrueNAS Scale. What doesn't sit well with me is having some agent in background burning CPU for what's to me no good reason. Wouldn't Docker cover most of my bases, running a bunch of crap like influxdb and mariadb in background, resource constrained? Correct, but some people either know or want to learn K8s which is fine. K3s works pretty well on a single node.
|
# ? Dec 8, 2021 22:48 |
|
Combat Pretzel posted:If this is a single host, not a cluster, do I even need something like Kubernetes? I'm "using" it because it came with TrueNAS Scale. What doesn't sit well with me is having some agent in background burning CPU for what's to me no good reason. Wouldn't Docker cover most of my bases, running a bunch of crap like influxdb and mariadb in background, resource constrained? Hard to say if 20% of a core is much without knowing anything about your hardware, but if you were running Docker you would have dockerd always running and spending some amount of CPU cycles - although likely less than k3s considering the latter does a lot more. TrueNAS SCALE settled on Kubernetes for containers - which makes sense IMO considering Docker's waning relevance and poor business decisions. I don't think you'll get less overhead by running Docker in a VM in SCALE, so unless you're planning on changing OS I'd try to become friends with k3s.
|
# ? Dec 9, 2021 00:09 |
|
It's incredible we've progressed from containers being good because they're lightweight and portable to just fuckin shipping entire single-node kubernetes stacks as a distribution mechanism.
|
# ? Dec 9, 2021 01:00 |
Combat Pretzel posted:I'm setting up my NAS with TrueNAS Scale, which uses Kubernetes. BlankSystemDaemon posted:kubernetes is made for hyperscalers that need massive scale-out orchestration Methanar posted:It's incredible we've progressed from containers being good because they're lightweight and portable to just fuckin shipping entire single-node kubernetes stacks as a distribution mechanism. BlankSystemDaemon fucked around with this message at 01:12 on Dec 9, 2021 |
|
# ? Dec 9, 2021 01:07 |
|
Methanar posted:It's incredible we've progressed from containers being good because they're lightweight and portable to just fuckin shipping entire single-node kubernetes stacks as a distribution mechanism. Lightweight compared to running a whole VM, you mean. Everything is relative.
|
# ? Dec 9, 2021 01:21 |
|
Yeah. K8s/K3s are great if your service is lightweight enough to be scaled that way. The average application is not. Ask me about trying to convince Devs not to roll entire Java apps in K8s and then rolling my eyes when they didn't understand why it was so slow.
|
# ? Dec 9, 2021 01:23 |
Keito posted:Lightweight compared to running a whole VM, you mean. Everything is relative. Service jails on FreeBSD, where you have a single binary executable and its configuration file(s) in a container, have been a thing since 1999.
|
|
# ? Dec 9, 2021 01:25 |
|
BlankSystemDaemon posted:If anything, since the introduction of containers on Linux, they've grown more heavy-weight rather than light-weight: Yeah and K8s is just the inevitable end of Jails: What if we made an entire Jail filled with jails all controlled by a central control pane. Also: What do you mean we have to update the images ourselves?!
|
# ? Dec 9, 2021 01:26 |
|
Yeah, well, I set up things directly in Docker instead of having Kubernetes gently caress around and set up Docker in background, and now the system uses gently caress all CPU power when it idles, unlike k3s which was just hugging 25-50% of a CPU core because Methanar posted:It's incredible we've progressed from containers being good because they're lightweight and portable to just fuckin shipping entire single-node kubernetes stacks as a distribution mechanism. This Kubernetes thing, at least outside of the alluded "hyperscaling setting", feels like someone's pulling a huge practical joke on me/us.
|
# ? Dec 9, 2021 03:00 |
|
Keito posted:Lightweight compared to running a whole VM, you mean. Everything is relative. You can use a hypervisor as your oci compliant 'container' runtime now for some reason.
|
# ? Dec 9, 2021 03:31 |
|
Combat Pretzel posted:If this is a single host, not a cluster, do I even need something like Kubernetes? lmao
|
# ? Dec 9, 2021 20:12 |
|
Combat Pretzel posted:Yeah, well, I set up things directly in Docker instead of having Kubernetes gently caress around and set up Docker in background, and now the system uses gently caress all CPU power when it idles, unlike k3s which was just hugging 25-50% of a CPU core because It's a poor man's Erlang.
|
# ? Dec 16, 2021 08:38 |
Mr Shiny Pants posted:It's a poor man's Erlang.
|
|
# ? Dec 16, 2021 09:00 |
|
Mr Shiny Pants posted:It's a poor man's Erlang. BlankSystemDaemon posted:That's the best description of kubernetes I've ever heard. I have told people roughly something like this before. I wrote a lot of Erlang once upon a time. When I first encountered k8s, my initial reaction was “huh, someone recreated erlang but shittier and with more steps.”
|
# ? Dec 16, 2021 11:45 |
|
I have a weird question on VMware SnS , can you reactivate an expired contract (say like dell server hardware support) or is it like microsoft software assurance where once you stop paying you cannot renew ever again?
|
# ? Dec 16, 2021 16:24 |
|
SlowBloke posted:I have a weird question on VMware SnS , can you reactivate an expired contract (say like dell server hardware support) or is it like microsoft software assurance where once you stop paying you cannot renew ever again? Yeah if you own the perpetual license but the support has expired on it, you can renew support on that license.
|
# ? Dec 16, 2021 17:05 |
|
Mr Shiny Pants posted:It's a poor man's Erlang. Either way, this Kubernetes control plane being wasteful as gently caress in regards to CPU cycles, why even? Everything is event driven in operating systems nowadays. If nothing or barely anything is happening, why is it burning that much CPU.
|
# ? Dec 16, 2021 17:07 |
|
Tev posted:Yeah if you own the perpetual license but the support has expired on it, you can renew support on that license. There's an extra fee though. They don't want you renewing just in time to open support tickets and then dropping again. $job found a license where someone had inappropriately dropped support years back, and when we asked VMware for a quote on getting it back under support, they told us that at this point it would be cheaper to buy new licenses.
|
# ? Dec 16, 2021 17:44 |
|
So I've given up on ESXi and trying out Proxmox now. One thing I can't figure out... in ESXi, you spawned vmkernel NICs for access to management interface. How does that work on Proxmox? I can't seem to find an obvious place to control where the management interface listens. It is a DeskMini 310 w/ single NIC I'm finally going to colo, so in my ideal world, I'd like too: Untagged <--> eno1 <--> vmbr0 <---> pfSense VM WAN Interface VLAN 10 <--> eno1.10 <--> vmbr0.10 <--> Proxmox management interface And then vmbr1 is Proxmox, pfSense LAN, TrueNAS and my Linux VM. This will stop me from getting locked out of Proxmox if pfSense is down / let me use it for initial setup at home.
|
# ? Jan 12, 2022 03:03 |
|
|
# ? May 9, 2024 15:33 |
|
Combat Pretzel posted:I don't know what that means. I know that it's a programming language, but that's the extent of it. Clustering physical machines and running microservices on top of them is Erlang in a nutshell. Erlang is just one VM (BEAM) that can run actors (microservices that pass messages) with supervisors (if you use OTP). You need to program in Erlang but from a technical standpoint they are roughly comparable at what they accomplish. Erlang is elegant though (like really elegant), Kubernetes is not IMHO. All that is old is new again. It was more of a reply to: This Kubernetes thing, at least outside of the alluded "hyperscaling setting", feels like someone's pulling a huge practical joke on me/us. I got that feeling as well. Mr Shiny Pants fucked around with this message at 08:50 on Jan 12, 2022 |
# ? Jan 12, 2022 08:39 |