os-net-config just sends a bunch of pings without waiting for anything in /usr/libexec/os-refresh-config/configure.d/20-os-net-config. There are some situations where a ping will fail instantly thus not leaving enough time for a port to come up.
Current code:
~~~
until ping -c 1 $METADATA_IP &> /dev/null; do
COUNT=$(( $COUNT + 1 ))
if [ $COUNT -eq 10 ]; then
echo "FAILURE"
echo "$METADATA_IP is not pingable." >&2
exit 1
fi
done
~~~
It could be like this:
~~~
until ping -c 1 -W 1 $METADATA_IP > /dev/null; do
COUNT=$(( $COUNT + 1 ))
if [ $COUNT -eq $METADATA_IP_PING_TIMEOUT ]; then
echo "FAILURE"
echo "$METADATA_IP is not pingable." >&2
exit 1
fi
done
~~~
That way, we are going to try 1 ping every second, for $METADATA_IP_PING_TIMEOUT and it's going to be possible to define that amount of time.
Steps to Reproduce:
1. Disable portfast on switches
2. overcloud deploy
3.
Actual results:
Overcloud deploy fails because the ports can take 15-30 seconds to come up
Expected results:
os-net-config should wait a bit more time before failing
I went ahead and submitted a patch https:/