My thoughts on buying a system with proprietary components…
Pros
Cons
Single host - Just Docker run + Portainer - Also using Macvlans so most containers have hostnames and static IPs on my LAN. K8s is cool, but I have no need for container orchestration.
Choosing a service to NOT selfhost is a subjective descision.
I host 18 Proxmox VMs and 20 Docker containers at home. I also was selfhosting a WebDAV server for synchronizing my Joplin notes between devices and Vaultwarden for managing my Bitwarden vault, but decided to push the Joplin synchronization target to Dropbox [free] and to use Bitwarden’s free cloud solution for my passwords and secure notes. I did this because I will need immediate access to these two critical sources of information should my house burn down, or get blown over by a tornado. I have extremely strong passcodes for these and trust the hosts.
This was strictly a personal decision. YMMV.
OP, your poll needs an “All of the above” option. That would be my choice.
In a nutshell, I have no interest in the feature bloat being introduced into Plex. I want a simple media server. Jellyfin provides this, and includes basic features like hardware transcoding and live-TV. I migrated from Plex [free] to Jellyfin a couple of years ago and haven’t looked back.
I also am in the Apple ecosystem and pay $10 USD per year for the Infuse app, which is more polished IMO than both the native Jellyfin and Plex clients.
The Tiny (Lenovo ThinkCentre and ThinkStation), Mini (HP EliteDesk and ProDesk), and Micro (Dell OptiPlex), also known as 1-liter PCs, are all relatively similar. ServeTheHome (STH) does a good job of comparing these in Introducing Project TinyMiniMicro Home Lab Revolution. They also have a YouTube channel where Patrick reviews and opens up selected models.
SFF PCs of the same series from these manufacturers are similar as well, but vary more in overall dimensions. HP and Lenovo SFFs have slightly larger footprints than the Dell OptiPlex SFF.
All SFFs and Tiny/Mini/Micro PCs from these manufacturers have tiers from entry-level to Pro/Enterprise. These tiers, and sometimes even build-time options (Dell?) determine the features, eg. vPro and Intel Active Management Technology (AMT) support, number of NVMe M.2 slots, and even a PCIe slot (Lenovo only) that the PC includes.
I know nothing about HP SFF and Mini PCs, but can offer some insight on Dell and Lenovo…
DELL:
Around 2015 Dell started using model numbers that represent the tier and generation. The tiers were 30x0, 50x0 and 70x0, where 30/50/70 represents low to high end and x0 represents the generation. For example, 3050, 5050 and 7050 can use 6^(th) or 7^(th) generation Core i CPUs. Some models ending in an odd number, like 5055, take AMD CPUs.
The Wikipedia Dell OptiPlex page is a good overview of Dell models.
Lenovo:
Lenovo has recently changed the scheme used for model numbers, so I am a bit out of date. Lenovo PCs that I am most familiar with are the M7x0 and M9x0 series from around the 2016 to 2019(?) era. The 7x0 series are entry level and do not support vPro/AMT. The 9x0 series support vPro/AMT. The x0 again represent generations, specifically what generation Core i CPUs are supported. The M710 and M910 can use 6th or 7th generation Core i CPUs. The M720 and M920 can use 8th or 9th generation Core i CPUs. Lenovo ThinkCentre model numbers also end in the letters q or x for Tiny PCs, and s for SFF PCs. The q or x designations represent additional features available. For example, the M910x has two M.2 slots and a PCIe slot. The M910q has one M.2 slot and no PCI slot. This changed in the next generation where the M720q and M920q still have only one M.2 slot, but also have a PCIe slot.
Lenovo also has ThinkStation (their workstation-class PCs) in a Tiny form factor, these are ThinkStation P310, P320 P330, etc.
If you’ve gotten this far, I am impressed. As you can see, the varying number of model for these are maddening. I suggest deciding on the features you want and then dive down the rabbit hole and research them. The manufactures (well Dell and Lenovo anyway) have good documentation and specs on line for their PCs.
Telegraf with Docker Input Plugin installed on the Host writing to InfluxDB and displayed in Grafana, both running in Docker containers.
Here is a screenshot of my Server Performance Grafana dashboard.
There are a lot of good responses here that I won’t reiterate. I will say that, in my own personal and professional experience, Linux simply is a far better server OS than any Windows OS.
That said, use what works for you. If you are experienced and comfortable with Windows, and aren’t too keen on climbing the Linux learning curve, then by all means use Windows.
My only suggestion would be to use Windows Pro (for RDP), or find a Server Standard license for sale at less than retail.
A server rack certainly is possible, but an appropriate furniture cabinet may function for housing your lab as well. It just needs to be deep enough for your equipment and well ventilated. This can even be custom built if you have woodworking skills and tools.
I built this cabinet a few years ago. It meet with my wife’s approval not because it was pretty, but it wasn’t too ugly.
Debian rules!
I use Homepage. There are plenty of other options, but I prefer Homepage because of it’s speed and simplicity. Many landing page apps allow for customizing on the page itself. Homepage is configured using YAML files, and therefore static, so family members will not be able to make changes, or accidentally break it, once you have it set up.
Here is a screenshot of my… Homepage.
I use custom images and icons uploaded to Dropbox in… icons-72x72-png.zip
These are kept in volumes for persistence…
--volume homepage_config:/app/config \
--volume homepage_icons:/app/public/icons
--volume homepage_images:/app/public/images \
It may be me, or Reddit could be flaking out, but it appears that everyone except you and I have deleted their comments. I also cannot reply to your comment about N8N, so I am commenting again. Weird…
I recommended Pushover in my previous comment. To address your initial question, I’m running Docker containers for Mailrise, which provides an SMTP gateway that embeds Apprise code, and also Apprise standalone. I use Mailrise for services that support only SMTP notifications (e.g. Proxmox). Mailrise translates an email message and, using Apprise code, submits it to Pushover. I use Apprise standalone for receiving notifications from Healthchecks, which monitors cron jobs. Healthchecks supports a number of notification services, but standalone Apprise was the best fit for me.
As I understand, Apprise has its own light-weight HTML API for creating and configuring notification endpoints. It can create a webhook endpoint, but I don’t believe it receives messages as webhooks.
Like me, you probably will need multiple services. I tossed up an N8N container out of curiosity. It can accept a webhook and then act on it. It also supports Pushover. So, it looks like a webhook to N8N can be translated to a Pushover notification. Using Mailrise (for SMTP messages) along with N8N (for webhook and other messages) should provide the extensibility you’re looking for.
Again, I highly recommend Pushover. Receiving notifications from many different sources, that are organized into “applications” with their own icons, in one mobile app is a tremendous benefit… to me anyway.
does it also report DHCP reservations?
Thanks, and yes, Type “static” are DHCP reservations.
Where do you do DHCP, on the PFSense or somewhere else?
Yes, on pfSense. I use the Python function written by pletch/scrape_pfsense_dhcp_leases.py (on Github) that scrapes the pfSense status_dhcp_leases.php page. Then added my own function for querying my TP-Link APs using SNMP to determine which AP a wireless DHCP client is connected to.
I can throw the script up on Dropbox if you are interested. I am mediocre at writing Python, so it is pretty specific to my environment.
I use Telegraf + InfluxDB + Grafana for monitoring my home network and systems. Grafana has a learning curve for building panels and dashboards, but is incredibly flexible. I use it for more than server performance. I have a dual-monitor “kiosk” (old Mac mini) in my office displaying two Grafana dashboards. These are:
Network/Power/Storage showing:
Server sensors and performance showing:
Netdata works really well for system performance for Linux and can be installed from the default repositories of major distributions.
I use Joplin every day for taking my own technical notes, often with screenshots. I use it across multiple devices and operating systems that sync through Dropbox. Notes are exported to PDF when there is a need to share them, which is infrequent.
Zotero looks keenly focused on managing research information and creating bibliographies. I am not actively involved in research, but I just downloaded and installed it on my M1 MacBook Air out of curiosity.
I opened the Zotero app, tried to create a new folder, and it immediately froze. It also looked outdated. Hope you have better luck with it.
That’s a lot of Linux ISOs.