It will please everyone that's invested that the UPS in my #homeprod that needed batteries ages ago now has fresh batteries, estimated runtime is now one hour 45 minutes (down from less than 10 minutes before it gave up during the last outage).
I asked the person who gave it to me, this UPS is *ten years old* now, and it sat for two of those years powered off and dead, all on the original batteries, so fair play Eaton, I think it's done pretty well.
* Because the power required is often higher for big machines for idle time.
* If one machine dies, assuming they are equally distributed, you only have to find places for 33% of your services and not 100% of services.
* Same goes for configuration braking changes, 33% or 100% down.
Do you have #homeprod ? How much downtime with room mates / life mates accept?
@tisba @jwildeboer @homelab_de
I see, if I were you I would play a bit with zfs on the side and only then migrate anything #homeprod in your #homelab :D
And for off-site backups you can do simple sync/incremental backups. Borg and restic sound good, there are also some others but from all the similarly named tools I keep forgetting which ones were unreliable and should be avoided.
Wrt ZFS the biggest thing for me is the ability to snapshot and then move the snapshots around, clone into a new dataset etc. I just had a major scare because my whole camera/gallery storage was gone (rsync reporting "created directory /mnt/path/to/photos"). In the end it was a case of the dataset not being mounted properly but all the daily/weekly snapshots meant that I havent lost anything from my previous phone for example:
- `zfs clone pool/dataset@snapshot pool/cloned-dataset` and verify that my files are there.
- `zfs rollback ...@snapshot` to roll back.
Or hell, make changes to a clone and promote it to replace the origin/primary. At zero extra storage cost.
So, enough preaching! But #zfs is really great.
I recently had a lot of issues with uptime, precisely since I moved flat
Turns out, supporting a 'big' Homeprod on a 5G link is not a very good idea
Especially because of latency, my remote Control Planes were getting kicked because healthchecks were taking too much time to complete
Buuuut now we are back to full blown Fiber, and a guaranteed fixed IP
Hopefully should remedy a lot of the previous issues I had pre-move related to uptime
#homelab #homeprod #selfhosted #networking #network #kubernetes #k3s #uptime #fiber #5g
My theory is that a single more powerful unit, while there might be some occasional downtime due to maintenance the significant reduction in complexity & increase in single host performance specs might balance out....
I like the two #intelN100 that I currently have but they feel a bit cheap for #selfhosting #homeprod
I would also consider something else, but I think I have already done too much research π
I need some help from the #selfhosting and #homelab community, I have reached full analysis paralysis!
I have 3 weeks left to determine what hardware to buy for my #selfhosted #homeprod my goal is to run #nextcloud & #immich in production well enough that It can seamlessly support my wife to #degoogle
I was leaning towards #talos cluster or #k3s for #kubernetes and #longhorn storage, but might all be overkill
Have determined I have these options are similar price & energy budget points:
Hydrian
fraggLe!
Boris Marinov
DocYeet 
Michael