Re: Is there a generic canonical way for a package script to check network connectivity?

[Thomas Goirand <thomas@goirand.fr>]

On 10/9/23 07:20, Simon Richter wrote:
> Hi,
>
> On 09.10.23 11:49, Jonathan Kamens wrote:
>
> > I need to be able to tell from one of my package scripts whether the 
> > host has networking connectivity.
>
> 無. There is no such thing as "networking connectivity."
>
> There is "has access to a particular service, at this precise moment in 
> time" and "is usually connected to an unmetered Internet connection so 
> there is a reasonable expectation that a particular service can be 
> reached and should be made part of a workflow."

I very much agree with Simon. There's all sorts of "networking 
connectivity", partial or not. For example, would you consider that your 
server has network connectivity if it only has L2 link local 
connectivity to the neighboring switches? In some case, you shouldn't. 
In some case (like using bgp-to-the-host setup), you should first check 
if FRR is using it to provide wider connectivity. In the case of an 
OpenStack cluster, in some setup, FRR would do that, but typically the 
nodes don't have internet access.

After many wrong designs, I ended up having a process that pings the 
service I need to access to, with a script like this one:

TRIALS=300
while ! timeout 2 ping -q -c 1 ${DST_SERVER} >/dev/null \
		&& [ "${TRIALS}" -gt 0 ] ; do
        sleep 1
        TRIALS=$(( ${TRIALS} - 1 ))
done

if [ "${TRIALS}" = 0 ] ; then
        exit 1
else
        exit 0
fi

Since I use this, I never had any issue.

(comments on the above script welcome... one improvement would be using 
a timestamp using date to write the timeout in seconds, but in this 
particular instance, I don't really care, I just care the caller script 
waits until there's connectivity)

Cheers,

Thomas Goirand (zigo)