Posts by chewitt

Yup, it's always easier (and considerably faster) to backup and restore the 1-2GB (average) of actual content to a freshly installed card than backup and restore 8GB of full card image. Users seem to like doing it the hard way though

The default KEYDB.cfg file we include is only a placeholder. You will need to source your own file that actually has content before discs will work. Then if you're lucky there are keys in the file for the discs you're trying to play.

Quote from shippy

Any ETA on WG becoming part of the Linux kernel, since Torvalds loves it so much?

Nope. And considering it didn't even compile on ARM hardware until I made an attempt (and failed, and reported the bug to the devs) in late December I'd guess WireGuard has some way to go before it gets merged. It's great but definitely a bit rough around the edges.

It shouldn't be too hard to figure out a systemd service that can be placed in /storage/.config/system.d/ to start the server.

"WARNING: [pool www] server reached pm.max_children setting (5), consider raising it"

^ that setting needs to be raised in the php-fpm config, but whether that's the cause or simply a symptom of simply running out of RAM is anyone's guess. For the quantity of containers you're running an RPi3 is inadequate. You'd be a lot better off with a basic NUC or a basic Synology NAS and then use the RPi3 as a dumb client device.

Seeing error messages like "Your kernel does not support cgroup memory limit" makes me think the Allwinner kernel defconfig doesn't have all the required cgroup/namespace options set.

Ping jernej ^

The geometry of the partition scheme and filesystems on the card to not match the geometry of the card. Fix that (and resize the /storage partition) using Gparted from an Ubuntu LiveUSB (or similar) and you'll probably solve the issue.

TVHeadend (server) add-on and the Kodi PVR (client) add-on for TVHeadend. Both are in the LE repo.

RPi Zero is going to be weak for this. RPi3B+ would be better.

WireGuard is a kernel module (currently out-of-tree) and a userspace config tool (wg) that are used to configure a wg0 network interface that presents in the OS much like a normal Ethernet interface. I'm still actively working on the userspace packaging so balbes150 images are usable with some caveats but also subject to change. I've also identified an interesting connman bug with NTP (reported upstream last night) that needs fixing before it'll be possible to use automatic WireGuard connections on boot. It's also untested on Generic/RPi so I need to spend some time on that too. I'm hoping to get it merged as an LE9.2 feature. Compared to OpenVPN the initial config process has more confusing terms (lots of things that sound the same) but it's actually rather simple and connections take a second to create.

Have a look at /storage/.config/wireguard.conf.sample .. rename to wireguard.conf and it will start the connection on reboot, but this will cause NTP to fail and if the system clock is wrong most SSL certificates will fail validation causing other issues. You need to stop, pause, restart wireguard.service to workaround that until the connman fix comes in.

project: add wireguard package · chewitt/LibreELEC.tv@d82a723 · GitHub

balbes150 images do not have the latest changes I pushed last night so "wg-quick" does not exist until his next rebase/release.

???

I've been experiencing the same "connman attempting to use IPv6 NTP servers" issue while poking WireGuard support. I finally managed to track down one of the connman developers, which led to this chat:

[22:28:39]  <wagi>    IIRC, we try resolve the nameservers and when we get an IPv6 we try it. Don't know if we have some logic in place which checks if IPv6 is up
[22:31:04]  <wagi>    Indeed, we try IPv6 without checking if IPv6 is setup
[22:31:20]  <wagi>    So that explains the 'Time request for server' message
[22:31:44]  <wagi>    We should check if IPv6 is up and running before trying to send IPv6 frames :D

Fingers crossed there will a patch to test in the near future

MikeKL my branch currently has some experimental code for WireGuard support and I modified /usr/bin/usercache-setup to create the WireGuard keys, so my guess is /usr/bin/wg is missing due to WireGuard support not being included in balbes150 images, and thus usercache-setup.service fails. If that's the case it's nothing serious and I'll tweak the script so it avoids the failure.

If you install default LE you get a 512MB boot partition which is approx 2x. the size needed for our normal image boot files. Nothing wrong with that.

RR needs considerably more space (about 700MB) which is why it's creator uses a 1024MB boot partition size in his install images.

So you all you need to do is install using the RR .img.gz that 5schatten provides.

NB: your KEYBOARD appears to have some caps lock PROBLEMS.

Gitlab is a bit more advanced in overall features than GitHub, but ultimately git (at the command-line, where all the real work gets done) is just git and unless you're needing to fiddle with CI pipelines I find them to be much the same; visual differences aside. My personal preference is GitHub due to more whitespace in the GUI and finger memory of how things work, but I'm still closer to n00b than power users so I wouldn't assign much gravitas to my opinion on the topic. Try both, whatever works for you and results in public shared code is a good thing.

c.2001 I used to heat the ground floor of my house in the UK with the thermal output of a NetApp filer cluster

Quote from relkai

btw.: Could we please stop calling this metal thingy a "radiator", but a heatsink instead?

To explain: balbes150 reads this forum (and replies) entirely via Google Translate, and while it does a reasonable job, it also makes some persistent mistakes including heatsink > radiator, compile > assembly, and banned > illegal. I've also noticed that Google frequently uses English words that are stronger in meaning than the original Russian text, so it's best to not read his words too literally. Once in a while Google also mangles his sentences into something unintelligible (even if you understand Russian it doesn't help) but that's all part of the fun contributing to a global project

Please keep the conversation civil or this and other threads will be locked and moved to the off-topic bin or be trashed. The N2 is a reasonable board and we (all) look forward to good open-source support for it and other next-gen Amlogic devices.

Yes, as it contains the contents of /storage/.cache .config and .kodi .. and autostart.sh lives in .config