I was eyeballing a MD3600 yesterday, for only 150$.

Went back and forth on the idea of running it for an iSCSI san… but, remembered why I prefer zfs and ceph over HW raid.

If you want 10G performance, you need to get a 10G nic. They are only 30-40$ on ebay.

While, you CAN bond a pair of 2.5GBe ports, and POTENTIALLY get 5g of throughput, it will not be on a single session. ie- you can’t download a file at 5Gbps.

10G hardware is cheap.

I don’t think homelabs were ever the intended audience. There are MUCH more price effective, reliable, and performant options over their cases + expanders.

Even if you do want to do casaOS, or linux- I’d still recommend putting proxmox as the base os.

I did put the disclaimer front and center! Ceph really needs a ton of hardware before it starts even comparing to normal storage solutions.

But, the damn reliability is outstanding.

https://static.xtremeownage.com/blog/2023/proxmox-building-a-ceph-cluster/

Having around 10 total enterprise NVMes, and 10G networking, I am pretty happy with the results.

It runs all of my VMs, kubernetes, etc, and doesn’t bottleneck.

Depending on the use-case, absolutely.

For a small site, absolutely.

I have a few dozen externally exposed projects that I self host though, a few of them are rather resource intensive, which would add up pretty quickly in AWS.

In my case, keeping everything in an isolated DMZ, handles reducing the risk vastly, as well as completely isolating internet-exposed applications from everything else.

It’s all about have proper redundancy, and risk-aversion.

And, of course, working backups, and a contingency plan when something bad happens.

But- then, how are my side-businesses supposed to make money?

Storefronts, and other externally-exposed services generally don’t work too well… when they aren’t exposed.

No- Cleantalk.

Do have auto-updates. Don’t have wordfence- just heard about it, yesterday.

But, I do have another service that provides vulnerability scanning… and after checking my emails, it has been trying to notify me of the issue for about a month now… Suppose, I need to actually check those.

Found a seller with Brocade ICX-6610-48Ps for 45$ each, and picked up a loaded MD1200 (24x2.5") for a touch over 100$.

Finally had my power usage SOMEWHAT under control… and now, I guess its time for the cycle to repeat itself…

I’m gonna have to add more solar panels to support this hobby. lol.

If, your hardware supports Intel vPRO / AMT (Half of my optiplexes did, as did my HP z240s)… That is a built-in hardware KVM.

MeshCommander can expose this is a nice way -> https://www.meshcommander.com/meshcommander

As a splunk architect- I really enjoy it.

For home use, its ok. But, without the enterprise features, it limits a lot of the capabilities.

You CAN use cribl.io with it, to replace a lot of the missing features… and to reduce the amount of data being stored. It has an extremely generous 1T/day free plan.

You can also use the universal forwarders, as they do not have a license attached.

Data is only licensed when it is written by an indexer.

There, are also ways of using the enterprise plan… by selectively not storing certain files under /etc… and restarting the container every few days.

I personally, use a bash script, which runs as a daemon, that dynamically sets the fan speed based on temps.

So, under idle, the servers run nearly silent. Under load, the fans can spin up.

When the script/daemon stops, it restores the default fan curve.

Dumb UPS + Smart Plugs (S31) / Smart Strip (HS300) = Smart UPS.

S31: https://static.xtremeownage.com/blog/2023/sonoff-s31-low-cost-energy-monitoring/

HS300: https://static.xtremeownage.com/blog/2022/kasa-powerstrip-as-pdu/

Yup. You can do that.

Although- you wouldn’t “import” your existing containers. but, you can…

1. Create manifests for your containers (Kubernetes runs the exact same docker containers). or, find helm charts for your containers.
2. Import the storage from docker into your new PV/PVCs.

I would, suggest learning kubernetes first though. Learning curve can be rather steep.

Also, rancher + k3s would work perfect for your Pis.

Well, I run my containers in kubernetes.

And, it more or less includes full support for prometheus/grafana/alertmanager/etc.

So- I use that.

I personally use a combination of…

Per Circuit Monitoring: https://static.xtremeownage.com/blog/2023/home-solar-project-part-3-monitoring/

Sonoff S31s for individual devices: https://static.xtremeownage.com/blog/2023/sonoff-s31-low-cost-energy-monitoring/

A few HS300 Power Strips: https://static.xtremeownage.com/blog/2022/kasa-powerstrip-as-pdu/

I honestly got tired of APC/CyberPowr units either taking a complete shit (internal failures), or yeeting batteries every year or two.

I ended up building my own damn UPS, a few years ago.

https://xtremeownage.com/2021/06/12/portable-2-4kwh-power-supply-ups/

I have tested with up to a 1,500 watts constant load, and it was able to power a 1.5kw load for around an hour and a half.

It should be good for 20 years or so. Expensive, but, damn, peace of mind is nice. Under typical load, it can power my entire rack from around 4-6 hours depending on load…