Thank you for your quick reply. I understand your frustration on systemd, and was completely surprised myself to find out my network interface wasn't called eth0 anymore.......
However I can try to rename it, but I control the server remotely over ssh, and don't want to break the connection off course, or I'll lock myself out. For the moment, I'm not considering to travel with keyboard and monitor to our sailing pavillion to sort this all out, and leave that as it is, I'm afraid. I'll dive into option number 2, when I have time for that. I'll keep you updated on that. But thank you for your effort, I anyway really appreciate your extensions to weewx! Eelco Op dinsdag 18 oktober 2016 20:02:33 UTC+2 schreef mwall: > > On Tuesday, October 18, 2016 at 1:25:14 PM UTC-4, Eelco F wrote: >> >> Since then, CPU_temp is no longer read out, and also neteth0rbytes and >> tbytes are now longer written to the graphs. I think the latter must be due >> to ubuntu which calls eth0 enp0s25 now. This is due to systemd. >> > > why must the little children at redhat break everything with their shiny > turd called systemd? i have many systems whose network interface will only > every be eth0, so i have absolutely no need for 'predictable network > interface naming', and i definitely do not want my init system to choose > the name of my network interfaces, any more than i want my init system to > give me binary logs or doing my dns or dhcp or any of the other things that > systemd seems to be compelled to overrun. > > not only that, sometimes it is actually desirable to be able to refer to > the network interface generically. > > it is hard enough to write code that works across multiple operating > systems. the last thing we need is clowns like the systemd developers > changing things arbitrarily, with no reliable backward compatibility. > > i understand what they are trying to do, but the way they do it is > disruptive in a very unhelpful way. > > sigh. > > > >> Is it possible to adapt the code from cmon, to sort this out? >> Or could I adapt the code in cmon.py myself? I don't know a lot about >> python, though. But i can see that eth0 is mentioned in cmon.py as network >> interface. >> > > cmon does need a proper update. it has a lot of platform-specific code, > and there's nothing wrong with that. but the implementation is a bit > brittle. at the very least, i have to figure out how systemd does things > and deal with yet another special case for that. > > you might figure out how to get 'classic' interface naming from systemd. > according to this, it should be possible: > > > https://www.freedesktop.org/wiki/Software/systemd/PredictableNetworkInterfaceNames/ > > modifying cmon is a bit tricky. > > cmon gets the network interface names dynamically from the operating > system, including eth0. if you look at the raw packets from cmon (run > weewxd directly), you should see net_xxx_* observation names. > > the problem with eth0 in particular is that the label 'eth0' is used in > the database schema for cmon. > > so you have at least two options: > > option 1: in the cmon code, map whatever name you get from systemd back to > eth0. > > option 2: rewrite your database schema, i.e., change the column names from > *_eth0_* to *_xxx_* where xxx is whatever name you're getting from systemd. > > option 2 will probably be much easier than option 1. > > m >
