All - After some follow-up private discussions, the issue is that redhat builds `/etc/frr/daemons` file has `watchfrr_enable=no` as default behavior. This caused systemd to not see the registration from FRR that the daemons were up and running. systemd would then restart frr in a vain attempt at making the problem better. systemd systems require watchfrr functionality if you are using the /etc/frr/daemons and tools/frr control files in order to work correctly. donald On Wed, May 16, 2018 at 6:27 AM, Donald Sharp <sharpd@cumulusnetworks.com> wrote:

Yeah -

Let's figure out why zebra is being killed/crashing.

`systemctl status frr` -> immediately after you see this issue Where are the log files going `show log`? Where ever it is going let's look in that file for clues, if in syslog, run journalctl -f in another window while you start FRR up and watch what it says( report here too! )

If I had to guess without seeing anything I wold suggest that you are using systemd integration without compiling FRR for systemd. Did you compile with `--enable-systemd=yes`?

doanld

On Wed, May 16, 2018 at 4:24 AM, Edvinas K <edvinas.email@gmail.com> wrote:

...

Hello,

i'm using Centos 7 and FRR 4.0 and i'm getting "closing connection to zebra because of an I/O error!" when doing something in vtysh terminal.

Any ways to prevent this ? Thanks

_______________________________________________ frog mailing list frog@lists.frrouting.org https://lists.frrouting.org/listinfo/frog