FS#859 - OpenVPN is beeing killed with "service openvpn stop" due to short timeout

[openwrt-bot]

helgadeville:

TP-LINK TL-MR-3220 v.2.x with LEDE 17.01.2 onboard, installed OpenVPN.

Connection opens successfully, however, on stopping with "service openvpn stop", logread shows:

Tue Jun 20 08:36:35 2017 daemon.info procd: Instance openvpn::instance1 pid 4084 not stopped on SIGTERM, sending SIGKILL instead
Tue Jun 20 08:36:35 2017 daemon.notice netifd: Network device 'tun0' link is down
Tue Jun 20 08:36:35 2017 daemon.notice netifd: Interface 'vyprvpn' has link connectivity loss

This causes OpenVPN scripts to abort.

If stopping openvpn with "killall -TERM openvpn", then it gets enough time to end in a nice way and all scripts are OK.

[openwrt-bot]

jow-:

How long does your openvpn take to properly shutdown? Do you have any complex teardown scripts or similar?

[openwrt-bot]

helgadeville:

It takes around 5 seconds, there are no complex teardown scripts, just one sending echo 0 > brightness to led.
I have resolved this problem on my device using a patch for /etc/init.d/openvpn script (attached), which adds term_timeout parameter to service instance, but this is a part of the problem: another problem is, that /lib/functions/procd.sh ignores term_timeout parameter, so I filed another bug report (see link) and added a small patch for procd.sh script.
Those 2 patches make my installation work as expected with no further workarounds.

[openwrt-bot]

jow-:

Right, I already staged a fix for term_timeout handling, but I wonder what a giod default tineout for openvpn would be. I think about 30s, that should cover most usecases.

[openwrt-bot]

helgadeville:

I agree. I gave it 15 secs, but in case the internet connection is poor, the disconnect process may take longer.