Currently I have a Microserver G8, mildly upgraded to feature a 1265Lv2, 16GB, an SSD boot drive and a P222 controller with 4x4TB WD Reds. This server pulls double duty as a general purpose NAS (media for Plex, cold storage for "stuff" and a backup target) and Plex server. It used to run Unifi controller and NVR software as well, but I've split those roles off for reliability's sake to dedicated appliances (cloud key and UVC-NVR) and I'm thinking I'll go one more and split standalone NAS from applications (plex, and whatever else) So I'm after a sanity check/commentary for the hardware side... - Take the 1265Lv2 out of the microserver and put the G1610T back in for the dedicated NAS box. - Put the 1265 it in an as-yet-unpurchased C204 board, pop it an as-yet-unpurchased 2U case, throw in a spare PSU and SSD(s) for a dedicated apps server And the software side: - For the NAS, currently running W2016 - what about taking out the RAID card and using a dedicated NAS OS? I don't want to sacrifice space, so how well does a low-power CPU like the G1610T deal with parity in any one of the dedicated storage OSes? The reason for removing the RAID card would be to swap in a 10Gb NIC, would the low-power CPU mean that anything over 1Gb is superfluous anyway? - For the server... what to use. I have Hyper-V 2016 already, I'm not sure I know linux well enough to rely on it. Suggestions? I'm about to undertake the painful task of stripping down and rebuilding the rack, so kind of want to figure out if I even should save a 2U slot somewhere for a separate server in the first instance (even if I don't do it straight away) or not bother. I also want to do it on the cheap, since I don't have any problems with the way things are, it would just be more technically correct to change it up, and potentially give me more options in the future for... stuff
Yeah that was a thought, but if the Celery would make it a non-starter, presumably even a 1220L would be a limiting factor compared to a dedicated RAID card (part statement, part question)? EDIT: Also I should have mentioned in the OP, the server lives on a UPS, so write caching ins't a problem.
I've got openmediavault running very happily on my G8. I'd suggest following the manual install process - i.e. installing Debian and then adding the openmediavault installer. This means you can install the Teamviewer plugin if you wish for remote access to the console. Installing from the omv ISO release doesn't allow this as it doesn't include the correct drivers/grub settings and I just didn't know enough about Linux to fix it. I'm not really sure what you're aiming to gain from splitting it down except more power consumption from more hardware? If the box is handling the workload, let it?
It's a good point, and why I'm still kind of on the fence - more power consumption, more hardware, more space, more complexity... all of that. If it ain't broke... The main reason is that I'd like to feel empowered to do more tinkering, but don't want to do it on my main NAS. Sure I can spin up HyperV on that now, I don't like the idea of doing it on my main storage device, and would feel better if the NAS was static in configuration. Looking to the future as well, I also plan to move to 10GbE soon-ish, and there's not currently any room in the server for a 10Gb NIC without pulling the RAID card and doing it in software, something that isn't feasible in W2016 (unless I just mirror and lose some more space)
Same boat as you a while back. (Similar?) So I decided on the following. HPE ML350e gen8 v2 (following spec) - E3-1240 v3 - dual psu - SSD boot - 4x storage drives - RDX back up (will explain this later) - Windows server 2016 standard - Virtual machine server 2012r2 - Virtual machine synology OS. The SSD has windows, Virtual machine etc installed. From here I can then allocate storage per system using networked drives through the virtual machines. RDX backup to cover the OS installs, VPN server, DNS test server etc etc. When 10Gb nic comes along I can just add the part to the server. There’s plenty of power on tap here as this is essentially an i7-4770k with the extra gubbins you get from a Xeon. The synology side uses 2 gb of ram, there other OS has access to what ever is left after the main server uses what it uses from the 32gb installed. All boxes ticked, one box. If I break an OS I just use the backup from my RDX after deleting out the old VM. Files never get lost as they sit on the storage on the active local OS and are just fed in as Virtual network drives into the virtual machines.
This is a good reason to make the change. You might want to see if this https://forums.bit-tech.net/index.php?threads/new-asrock-rack-c2750d4i-and-16gb-ecc-ddr3.345425/ is still going? You could either team the 2x1GB NICs in it for more speed and leverage the extra sockets, with your RAID card in the PCI-E x8 slot, or you could use openmediavault's software RAID features (RAID0,1,5,6 and JBOD supported) and a 10G NIC, but the extra SATA slots onboard will mean you can add mooooooooore disks.
The urge to tinker is why I started putting everything on a dedicated server in the first place, and then I started to tinker with that. I feel your pain man. What applications do you want to run on the dedicated server? I've got plex server working brilliantly on both a Debian server and a Linux Mint desktop. I've got 3 unused rack cases, lots of hardware and a good working linux knowledge if I can help at all!
After familiarising myself a bit more with unRAID and FreeNAS I've decided on a slightly different path, and to simply pick up a NAS appliance such as a Synology RS1219+ and possibly use the microserver as an app server, but more likely, mainly for packaging reasons, replace it with a 2U box (thus freeing up a fairly valuable 2U in the rack). The server is currently running Plex, Roon and PRTG - nothing especially heavyweight (although 4K>1080 plex transcodes do max out the CPU) - but with more separation I'd be more inclined to do more with it.
