I had not figured out yet the optimal approach to upgrades as it is really terribly painful procedure. First you have to get sources from git, apply LXC patches, compile whole distribution and solve the compilation issues as very rarely it go flawlessly, then transfer everything rootfs, kmods, kernel to target device and unpack to rootfs copy new kernel, boot it and reinstall all the IPK packages and finally transfer /etc/config, reboot and reconstruct from debris what is broken and test all services one by one and eventually adjust configuration in case something changed. For our current Turrises all of this mammut work is done by turris NIC team. But to make it yourself it is very different story as you constantly end up how to solve some problems and sometimes ending up with no answer at all. Thatās the reality of openwrt. If it works itās great, if not, it is pain.
Hence I considered also use combined-ext4.img.gz and put everything else on extroot but it has also some pro and cons.
There is not, you are right. But with that setup you are future proof and can easily downgrade to mPCIe as there are adapters available (the other way around is speedwise limited).
We should not talk about the details - those should be discussed in OpenWrt forum. But I think it is necessary for others to see there are options available, if one is a real poweruser willing to pay some more Euros/Dollars on ones home network.
Normal networking would be in my eyes
internet connection speeds up to 400MBit up/down if you take some time for tweaking; 1GBE wired and maybe 0,5GBE wireless network speeds, in this category Turris Omnia can serve very well as a router.
It can even do its job as a really small server - you can run some smaller containers for apps like pihole or nextcloud, it already has more power than my aged Synology DS213+.
Adding gadgets like BigClown devices you can build yourself a home automation centre
So Turris Omnia is a nice device.
MOXā¦ should have been neither advertised as smart home center nor nextcloud center - thats the job TO is fulfilling. MOX should have been build according to the requirements I stated above to just serve as a ap - trying to make it a allrounder like TO is in my eyes a design failure, the team should have focused on the network abilities, those really count and do have unfortunately bad teasing troubles
I hope the will succeed with the troubleshooting and MOX can sometime serve as a good ap.
But the downside, serveral guys are running into, is, that OpenWrt is not a distribution made for servers. Therefore it is a bad opinion for trying to run complex environments - @twinkie vividly discriped the problems that might occur when trying things offroad. That definitely applies for Turris as well - if you want to do things aside what I descriped above, run a distinct x86 server with a server OS (Red Hat, Ubuntu Server, etc) and host vms on it.
Turris Omnia/MOX might be expensive looking - but thats only on first sight (Example: ca. 800ā¬ for perfectly serving Wi-Fi for 150mĀ² on two floors with 1 TO and 1 MOX and some external antenna stuff). So this is perfectly enough for a network beginner and to learn.
An update to āindustrial gradeā (thanks @anon50890781, that sounds nice haha ) will add up to this 2 managed 10GBE switches and 1 managed 1GBE switch and a distinct x86 Server (10GBE, low power Intel Xeon, 64GB ECC RAM, 8TB PCIe-SSD with enough storage, 3 10TB external backup drives) - for some 1k Eurosā¦
In principle - yes. But as long as it is not possible to run unprivileged containers I wouldnāt expose them to the publicā¦
And I wouldnāt want to try complex things - but I stated this already above.
Sorry, but the initial idea is wrong, I think.
IIRC, the reason that CZ.NIC started with Turris 1.0 in the first place was to provide a mesh of probes that would collect data about traffic patterns on the Internet. And they wanted to get these probes everywhereādifferent providers, different locations etc., i.e. where the regular home CPEs are. Thatās why they were āgivingā routers to regular people. And they were thinking about adaptive firewall which would be updated from a central location based on the data that the probes were collecting. Right?
Then they came up with another ideaāa powerful home/SOHO router+switch+AP which would have the capability of: traffic data collection, adaptive firewall, honeypot (later HaaS), trivial web UI for non-tech home users, long-term software updates availabilityā¦ And weāve got the Omnia. Itās not a cheap device, I do not regularly run much services on it, but Iād keep it just for the software updatesāthe rest of features are just a bonus for me, I can tell you that!
Mox, yes, I have it, but I do not have any use for it. Maybe Iāll replace manageable D-Link switch with it, but itād be a waste. But again, software updates, the switch hasnāt got for years now, that may be the reason to use Moxā¦
Well, to concludeāI do not think we need a new device designed by this team, certainly not with the specs in the first post. Iād rather have the Omnia (and maybe Mox) manufactured for a couple more years and fullfill the other roles with a dedicated server/NAS/switch.
E.g.:
ethernet: If I need to run 2.5Gbps or even 10, 40 Gbps, Iād rather buy a dedicated device. My uplink is still symmetrical 50 or 100Mbps at this moment. I even do not know, because itās not a bottleneck.
wireless: if thereās a support in software, you can install an mPCIe card and connect it to whatever antenna you like.
storage/SATA: umm, no, thanks. I have an SSD inside my Omnia and itās OK. Then I have 2 NASes with 18 TB total. How big would be the power adapter if they run from Omnia?
expansion slots: 3 slots are not enough? Yes, sometimes, youāre rightā¦ If you care, but a full ATX size mainboard, please, and donāt try to put more into a CPE.
Itās true that (almost) anything is possible with a modular device. But Iād keep it for LTE, 5G, xDSL, fibre, etc.
Thereās one idea I may like to have: SO-DIMM slots for RAM. But I can certainly live without it.
Other than this, I pray for Turris OS to be released for another HW, e.g. Raspberry Pis. Iād have a use for this. If not, OK. A package with the adaptive firewall for arm, arm64, x86, x86_64 would be enough.
) will add up to this 2 managed 10GBE switches and 1 managed 1GBE switch and a distinct x86 Server (10GBE, low power Intel Xeon, 64GB ECC RAM, 8TB PCIe-SSD with enough storage, 3 10TB external backup drives) - for some 1k Eurosā¦