i wouldnt mind seeing wendell go in depth on the l1enterprise channel
I would love to see this. Knowledge of general networking is a huge help even when running home servers.
I have been running FreeNAS for a couple of years now. I think it solved a dream I had once, back in the early days, of having a computer that contained a library of every tv show I wanted to see. Well, the internet has also solved that problem, but nonetheless I am still happy with the media setup I have.
I am not running anything cool or underground, just all the usual stuff. A browse of linuxserver.io is a good start to see what popular things people are running.
I run Plex because I use a PS4 as my media PC. It has a native plex client. It is not ideal. If the internet is down, I cannot open it. I’d have to revert back to the smart tv app. I could try and use the browser, it has html5 video support. I live in Australia, so watching my stuff remotely is a laughable proposition.
Sonarr, Radarr, Headphones on the PVR side.
Sabnzbd, nazhydra, deluge, jackett on the download aggregation side.
Syncthing to sync data between various devices, and back to the server.
Things I haven’t found: A good music server/manager. Something with the power of media monkey. I have tried all the sonics, bliss, and a few others. None really do it for me. I even entertained the thought of running media monkey in a windows VM on the server, and access it via a web interface plugin, or over vnc. It worked, but didn’t have the utility the main interface had. I scrapped it eventually.
I have used other programs in the past. I will fire them up again when FreeNAS 11.1 rolls out.
Gogs for a git repository with web interface. Easy to set up, works well.
This isn’t going into pfsense, which you have covered elsewhere, or into Home Assistant, which has some crazy potential.
Things I thinks would be cool is Syslog servers / Radius servers / ips/ids ect. These are all things I am in the process of doing when I actually get the time. Right now just switched to ESXI from proxmox and run a Win16 server (free) Win 8.1 embedded and plex server on centos.
I would love seeing this kind of content too.
10G is worth it for ISCSI as it has several orders of magnitude more IOPS than gigabit Ethernet which ISCSI definitely benefits from.
This is something I rather enjoy.
Squid is a caching proxy for the Web supporting HTTP, HTTPS, FTP, and more. It reduces bandwidth and improves response times by caching and reusing frequently-requested web pages. Squid has extensive access controls and makes a great server accelerator. It runs on most available operating systems, including Windows and is licensed under the GNU GPL.
This is gold for anyone with a data cap.
Watching an into on the black magic that is Backula would be cool.
Backula and AMANDA are outdated.
If you’re interested in that sort of backup, check this out:
[Mod censoring political stuff]
Anyways, now that’s over, let’s get back on track.
I’ve compiled a short list of things that might be helpful to cover:
- docuwiki (I use it as a FOSS alternative for Confluence)
- Phabricator (Evaluating it as an alternative to Docuwiki and Gitlab)
- Kolab (Community)
- nextcloud/spreed (side hustle uses this for meetings and it’s more flawless than I feared, less flawless than I hoped)
This isn’t exactly home server per say, because I run a small side hustle where I provide FOSS/Linux solutions and support for small business.
I’ve been waiting for a video like this!
Here’s my setup:
5x4TB HGST Desktar Nas drives in RaidZ-2 for Data
1 Sandisk SSD for Jails
FreeNas 11 //duh
Emby //Plex Alternative that allows for custom CSS, better Kodi Plugins, etc.
SickRage, CouchPotato, Transmission
NextCloud //for Textbooks, Class notes, Homework, etc.
VM Container Running Ubuntu Server //LEMP stack for personal website
I am also currently using the file server to make offline copies of Wikipedia and StackExchange via Kiwix, as well as create local repositories for Fedora and Ubuntu.
My Future goal is to build a Threadripper Virtualization server for all the actual applications (emby, nextcloud, etc.) and just have my data on the current server shared via NFS, but that costs money…
[Mod censoring political stuff]
+1 for private git hosting (gogs/gitea/gitlab/phabricator)
PXE setup for booting installable linux and windows images (TFTP + HTTP + Samba), this is super useful if you’re doing lots of work in VMs.
And this is an example of why debate class in grade school should be mandatory.
My home server setup: Two ESXi machines
- ESXi Machine 1:
– CentOS 7: Plex server w/o external access.
– CentOS 7: LAMP Stack w\ NextCloud server that faces the internet and also acts as a proxy for the plex server.
- ESXi Machine 2:
– Windows Server: for games like Killing Floor 2
– CentOS7: Minecraft server for when I am bored
- Standard Router
– With a custom HOST file redirecting things like my nextcloud and plex back to local ips rather than use the external IP I’ve set on my domain name hoster.
- Seagate RAID 1 8TB NAS
- Windows Machine: This machine acts as an app server for shared application in the house.
You know, I am antsy around parity RAID. For varying reasons that include a $15,000 data recovery bill.
But then I found this article when trying to find the above linked one:
Checking your drives, yours are 1 URE every 10^15 bits, which lowers the chances of failure drastically.
“Why mention this when RAID Z-2 is essentially RAID 6?”
Because this problem applies to all parity RAID. Even RAID 6 is sketchy with 1 URE in 10^14 bit drives when your drives are 4TB or larger.
Personally, I’m a fan of RAID 10. Yes, you lose half the storage, but you get double write performance and quadruple read performance (if your OS supports it and Linux/FreeNAS do) while maintaining sub-RAID 6 availability but over RAID-5 availability. Sub-RAID 6 because you can technically have a failure of the array if 2 drives die, but it matters which two drives. Over RAID-5 because your RAID can survive if 2 drives die, but it matters which two drives.
You also get better server performance. Doing basic mirroring and splitting is easier on a computer than calculating parity bits constantly on writes.
Except for the fact that most websites that you frequently use are served over HTTPS which by itself cannot be cached unless your proxy is MITMing your connection to the server, which might not be possible, if the website uses techniques like HPKP.
Even if it does work for your use case you are putting the confidentiality of your transmitted data into the hands of the proxy server, so you need to make sure that it stays up-to-date with the current state of web security (e.g. what CAs to trust, what ciphers are fine, etc.).
I’m a little confused by this comment, but are you saying that my drives have a chance of error that is 1/10^15, which would be better than 1/10^14, but I should still consider mirroring for redundancy? Also, don’t mirroring and RaidZ perform differently based on the workload? Like one has better for async read/writes and the other is better for sync? (I may be way off base here but I think I remember reading that somewhere…)
I’m saying there are pros to mirroring and striping VS parity aside from just redundancy and failure rate.
And yes, 1/10^15 is better than 1/10^14. If I were you, I wouldn’t worry about your RAID failing, then being unrecoverable due to UREs. It is such a low chance of happening that you should be fine.
I actually have no idea regarding that one.
@sgtawesomesauce Do you know if this is true?
That’s a good question.
What I know, definitively, is that ZFS doesn’t perform well with synchronous writes. You’re going to get bad performance on that any way you slice it. It has to do with the abstraction layers and the way ZFS technically has multiple disk writes for each cluster write. (In a simple config, no mirroring or RaidZ, it would first hit ram, then it writes to the ZIL, then it writes to the final resting place of the data, long term storage on the drives)
I can only imagine that raidZ would be worse than mirroring considering it has to calculate parity as well and considering mirrors are usually striped together, meaning if you have a 6 drive striped mirror (raid10), you’d have the effective throughput of 3 of those drives, as opposed to the effective throughput of one in a raidZx.
To be fair, it doesn’t perform any worse than most (if not all) other solutions I can think of right now. But what you can do with ZFS is add a separate drive (preferably an SSD) as a ZIL. This device will act as a cache between the actual array and the rest of the system and will definitely speed up the IOPS for synchronous access.
When talking about synchronous performance, throughput doesn’t matter as much as raw IOPS. The latter doesn’t really scale on a mirrored array. However, I could imagine that a RAIDz array is actually slightly better if you are talking about 4+2 vs. 3+3, since you can sync different data to 4 drives at a time as opposed to only 3, but unless you are really hammering the array, I don’t think you can really feel that difference.