On Fri, May 05, 2017 at 05:39:24PM -0400, Michael Stone wrote: Control: tags -1 unreproducible
> Package: openvpn > Version: 2.4.0-4 > Severity: normal > > The systemd unit files for openvpn set LimitNPROC=10. This process limit > applies to all processes started by a particular unit, so (for example) if > there are 11 server .conf files the last one won't run. Or, with a smaller > number of .conf files, subprocesses (scripts) could randomly fail. There is a > note about this in README.Debian, but that note doesn't mention a likely error > message associated with nproc issues (generally something about a fork error). > I'd suggest adding a mention of fork errors as a symptom of the nproc limit > being too low, but I'd also suggest a much higher default limit. It is not > clear to me what problem a limit of 10 solves that wouldn't be solved by a > limit of 1000, while a higher limit is much less likely to be hit through > legitimate use. The LimitNPROC=10 definition comes from the upstream units, i.e. https://github.com/OpenVPN/openvpn/blob/master/distro/systemd/openvpn-client%40.service.in I'm not exactly sure why it was chosen this low, but I cannot reproduce your issue. I think LimitNPROC=10 is _per_ _instance_. I could start 20 server instances just fine for i in `seq 1 20`; do echo -e "dev tun\nifconfig 10.0.$i.1 10.0.$i.2\nsecret static.key\nport 200$i" > server$i.conf systemctl start openvpn@server$i done # pgrep openvpn | wc -l 20 Bernhard
