It's kinda mesmerizing watching the core speeds on one of the i7-12700Ts in the #Proxmox Cluster (the one that runs Tiggi.es presently) dynamically shift cores anywhere from 1400Mhz to 4.62GHz quickly, and on an as-needed basis. The low base speed saves energy, but it'll instantly pop cores up to speeds within its thermal/power limit envelope, so for lightly-threaded workloads, it can perform as well as beefier chips. #homelab#servers#cpu
@mario@stooovie I've looked at this setting and have tried changing it.
On Comet Lake and later, I'm seeing powersave disabling turbo boost and limiting cores to the base frequency while allowing cores to go even lower than base frequency before returning to it. Performance mode runs the cores at base frequency and permits them to turbo boost before returning to base frequency.
Given this behavior, and the low power nature of these systems (and the fact I've got solar panels generating an average of 50-60kW daily), I'll be leaving the systems on Performance Mode.
If I wanted to control turbo boosting, I'd just use powercap-utils and set the short-term power limit down, so the CPU uses less power when demand is there, but this will also impact performance negatively in high utilization bursty scenarios.
Instead of doing that, since I value performance over saving a few watts, I'm going to set my long-term power limits (thus TDP) up from 65w to 85w as the CPU cooler can handle it based on XTU testing, and short term will actually be lowered from 135w to 125w, but the window extended to ~57 seconds to enable larger workloads to complete more rapidly.
I can see Powersave making sense where Proxmox is being used locally with sipping power taking precedence over performance...but I bought these CPUs to perform, so I'm gonna help them do just that. ;)
I have a single-node #Proxmox setup with various services in containers and a #HomeAssistant VM. For example, https://blog.randomplace.online/ is hosted there. I needed a #Widnows build machine to build a work project, so I decided to spin up a new VM.
I got 10 minutes of build time and my server didn't die during the build. You can see the CPU load of two builds. There is plenty compute power left to serve my blog for that one visitor per month. Also, the Home Assistant VM was not affected.
Need to mention that the node is an #ASUS PN41 with 4-core Celeron N5100 and 16GB of RAM.
@estevez There's a lot of power to be squeezed from small nodes :)
My smallest one is dual core, four threads, running among others a Home Assistant VM and a Plex one - with bulk storage attached via iSCSI. No performance issues with any of the services on there.
@tuxedocomputers What would you recommend as HW for running #Proxmox as an enthusiastic developer for personal educational use? Will it perhaps be a supported OS during ordering?
I have a #proxmox problem where it looks like id's in the webgui are wrong.
If I connect to the command prompt it connects to the wrong vm. #homelab
Anybody see that problem before?
And know what to do perhaps?
Happens on #alpine vm's without kvm-client running
@ScaredyCat I have copied the disks and all is well. I think I was just too tired yesterday and cloned the wrong vm's. Its the best explanation I can come up with
My #Proxmox journey started a month ago. I moved all my all my #docker containers and #HomeAssistant VM from my #Synology NAS to a testing MiniPC with Proxmox.
This test was very positive, so I decided to move to a new server. I'm using a cluster to move my LXCs and VMs. I'm very impressed by how simple the process is for a newbie like me. I just need to remove my snapshots before.
Hrm, I think I'm calling it a day (or a night). Couldn't figure out yet why on this #Proxmox server a Linux bridge wouldn't forward #Multicast Router Discovery (#MRD) messages generated via @troglobit's mrdisc. While sending to ff02::6a via ICMPv6 echo request or UDP works just fine...
Disabling multicast snooping on the bridge does not help either. Nor does "ebtables -I FORWARD -p IPv6 --ip6-destination ff02::6a -j ACCEPT".
Likely sth. with the #ip6tables with nf-call-ip6tables enabled...
This evening was totally nuts. I checked, documented, and updated the firmwares on every SSD in the house. I took the entire house offline for nearly 4 hours. The kind of offline that means even DHCP wasn’t available.
Upshot. My #proxmox cluster lost quorum (because everything was off) and when I came to bring back up my core network services node it took me about 30 mins to figure out a) why are you not reachable on the network? b) how can I see your console (headless rack unit btw).
That node was forcefully removed from the cluster after some cursed action with an hdmi capture card to figure things out.
I’m almost at the point where I’ve finished my hardware stock take and can feasibly consider starting the Epyc build tomorrow or maybe Weds. #homelab
@ironicbadger I think the last time I updated a SSD firmware was for those old Samsung drives that didn't housekeep (840 Evo?) or the even older Crucial M4 drives that locked up after a few thousand hours. Certainly not a regular thing to do tbh.
Ugh, FINALLY got around to getting the last few drives installed in the new rackmount chassis. Figured while I was at it to go ahead and swap out for a beefy #ryzen7 with integrated #radeon graphics since it's a slim build! 😎
#proxmox is up and running, and before long, the #homelab will have another domain controller and some other dedicated #dev machines! 🙌🤘💯 :nixos: :windows: #server#programming
Well...my #proxmox server has decided it's going to be in a seemingly endless boot loop cycle. On reboot 4 in the last 30 minutes.
All I know right now is all of the boot on startup guest VM's start up and a few moments later we get an unexpected reboot.
The best part of course is that I haven't modified any guest vm's or really any config on the box in weeks. This is gonna be a problem for the me of tomorrow.
#HomeLab people, what are your favorite linux debugging/troubleshooting tools?
4 Lenovo Thincentres with a Protectli box all in a #ProxMox cluster, and a #Synology NAS for all the files to be backed up to. Each box has 250gb SSDs so they can manage most of their own stuff, but the NAS helps for HA and backups. The NAS does a great job hosting my photos and media, and the other boxes do great at.... Everything else!
@ironicbadger I thought the same thing and it is in the works! Currently modeling the first iteration and testing out the first print for scale, which was just barely too small. Printing the second test now.
After the size all goes well then I will make a thicker rack with spots for the power supply's, and then additions for the switch.
Once all that goes well I'm hoping to expand to the whole NAS et all!
@box464 for sure! So far I'm liking it. It's just an abstraction layer over docker, but it's nice. I'm getting ready to also install Homepage and see how that is compared to CasaOS. It's a little more Yaml config, but it does have widgets to get data from things like ProxMox and PiHole.
Enjoy and keep posting about the journey! I'm trying to and meeting some great people along the way.